ProgrammableMarketing/runanywhere.ai (opens in a new tab)
PDF

Views

  • Plan
  • Roadmap
  • Research

On this page

  • The brief
  • Start here
  • Idea portfolio
  • Effort vs impact
  • This week · 8
  • This quarter · 12
  • Big bets · 3
  • Wildcards · 1
  • How we can help
Generated Nov 10, 2026 1 view
  1. Programmable
  2. Playbooks
  3. RunAnywhere
RunAnywhere logo

growth playbook · Nov 10, 2026

runanywhere.ai (opens in a new tab)

RunAnywhere growth playbook

Diagnosis

RunAnywhere is a five-person YC W26 team that went from on-device SDKs (10.3k GitHub stars) to a hosted product, Wally, in September 2026, and now has ten open models, low per-token prices and a chart claiming about 1/23 of a closed model's cost per task. The product is ahead of the funnel.

runanywhere.ai has no prices and no pricing page (/pricing returns 404). The $5 starting credit is mentioned only in the refunds policy. The docs have no pricing or rate-limit page. The SDK audience (tutorials, README) is never pointed to Wally. The Apple engine has two names, and the newsletter archive on the main domain stopped at week 33 while the newsletter itself kept going.

This playbook first makes the free start and the prices visible wherever developers arrive, then turns the SDK and content audience into Wally users, adds model pages, proof and billing that keeps agent loops running, and finally bets on hybrid device-plus-cloud routing, dedicated boxes and an access-grant programme that fits the mission.

Growth Score

6/10

Biggest Opportunity

Developers already arrive through GitHub, the blog, the newsletter and 19 integrations. Showing them the $5 start, the price per model and the cost-per-task proof at that moment, and linking the on-device audience to Wally, could lift sign-ups and paid usage with very little engineering.

Biggest Risk

Hosted open-model inference is crowded: OpenRouter, Together AI and Fireworks sell the same models with bigger catalogues, customer proof and caching. A five-person team spread across hosted inference, two NPU engines and six SDKs may not get any one funnel right.

01

Start here

The three moves to make first. Each one is written up in full below.

  1. 1

    Tell developers about the $5 starting credit where they click "Get started": today it appears only in the refunds policy

    Add the $5 line under the homepage "Get started" button and at the top of runwally.com/pricing.

    High impact First result in 2 weeks
  2. 2

    Fix the /pricing 404 on runanywhere.ai and give the docs a pricing and limits page that coding agents can read

    Add a redirect from runanywhere.ai/pricing to runwally.com/pricing.

    High impact First result in 1 week
  3. 3

    Put Wally's cost-per-task chart on runanywhere.ai, where the GitHub and blog traffic lands

    Export the chart from runwally.com as a static image with source and date.

    High impact First result in 2 weeks

02

Your idea portfolio

24 ideas, grouped by when to run them. Open any idea for how to run it: the steps, who to approach, copy you can use, and how to measure it.

  • 7 ideas are marked Unexpected: plays you are unlikely to have tried.

Every idea placed by the effort it takes against the impact it could have. Quick wins sit top left; select a number to open that idea. The filters below narrow the list by status, timing, category and what Programmable can build.

Effort vs impactEach number is an idea. Select one to open it.
Impact →
Quick wins · 4Big bets · 7Fill-ins · 5Reconsider · 8
Effort →

Showing 24 of 24 ideas

This week 8

Quick wins with a clear first action.

  • 01

    Every new Wally account gets a $5 welcome grant, but the only place that says so is the refunds policy ("The $5 welcome grant is promotional credit, not a purchase"). The runanywhere.ai homepage has no price or dollar amount at all and says usage is "paid from prepaid credits you buy in the console". The runwally.com pricing page says "Prepaid credits, spent per million tokens" and never uses the words free, welcome or $5. The quickstart says only "Cloud usage bills against prepaid credits". So a developer arriving from GitHub, the blog or an integration reads "buy credits first" and may leave before discovering they can start for free. At $0.50 per million output tokens, $5 covers about 10 million GLM-5.3 Flash output tokens, enough for many real agent sessions.

    High impactLow effort First result in 2 weeks Co founder and CEO

    First action

    Add the $5 line under the homepage "Get started" button and at the top of runwally.com/pricing.

    How to measure it

    Visit-to-sign-up conversion

    Now
    –
    Target
    +30%
    By
    45 days

    Steps

    1. 1

      Agree the line

      One sentence on the $5 grant and what it buys in tokens, plus whether a card is needed.

      Co-founder and CEO 1 hour

    2. 2

      Homepage and Wally hero

      Add it under "Get started" on runanywhere.ai and runwally.com.

      Co-founder and CTO 2 hours

    3. 3

      Pricing FAQ

      Free credit, expiry and running out, using the refunds policy wording.

      Co-founder and CEO 2 hours

    4. 4

      Docs and CLI

      Quickstart step 2 and the message after `wally account login` show the balance.

      Co-founder and CTO 1 day

    5. 5

      Track the funnel

      Visits, sign-ups and first requests by source, for two weeks before and after.

      Co-founder and CTO 1 day

    6. 6

      Share it

      Post the change on X, Discord and in the next Inference Radar issue.

      Developer Relations & Technical Content Intern 1 day

    First result
    2 weeks
    Cost
    $0 beyond staff time; the grant already exists
    Owner
    Co founder and CEO

    Smallest test

    Add the line under "Get started" on the homepage only and compare sign-ups for two weeks.

    Why it works

    Write one line and use it everywhere a developer decides whether to try Wally: under "Get started" on runanywhere.ai, in the runwally.com hero and at the top of /pricing, in step 2 of the docs quickstart, and in the CLI's message after `wally account login`. For example: "New accounts start with $5 of credit: about 10 million GLM-5.3 Flash output tokens. Top up only when you need more." Add a three-question FAQ to the pricing page (Is there free credit? Do credits expire? What happens when I run out?) using wording already in the refunds policy, which says purchased credit does not expire while the account is open. Say plainly whether a card is needed to claim the $5. Then measure visit-to-sign-up and sign-up-to-first-request before and after.

    A free start with a concrete amount removes the biggest question a developer has before signing up, and the offer already exists, so the cost is a few lines of copy. Showing what $5 buys in tokens turns low per-token prices into something people can picture.

    Ready to use

    CopyStarting-credit copy

    Hero line: "New accounts start with $5 of credit: about 10 million GLM-5.3 Flash output tokens. Top up only when you need more."
    
    Pricing FAQ
    Is there free credit? Yes. Every new account gets a $5 welcome grant.
    Do credits expire? Credit you buy does not expire while your account is open.
    What happens when I run out? Requests return insufficient_credit until you top up in the console.
    
    CLI after login: "Signed in. Balance: $5.00 welcome credit. Try: wally opencode --cloud -m glm-5.3-flash"

    Risks and how to handle them

    • More sign-ups that only use the free credit

      Measure the share that top up within 30 days, and show the balance in the CLI so heavy users see when to buy.

    • Abuse of the welcome grant

      Keep the existing sign-in with Google or GitHub, and limit the grant to one per account as today.

    Evidence

    • Your websiteRefunds policy: "The $5 welcome grant is promotional credit, not a purchase, and is not refundable." Purchased credit "does not expire while your account is open"; prepaid Wally credits bought through Stripe are non-refundable except for billing errors, duplicate charges and platform faults.Source
    • Your websiteHomepage (checked November 9, 2026): H1 "We make the best open models faster and cheaper to run, for everyone." The hero button "Get started" goes straight to the console; usage is "paid from prepaid credits you buy in the console"; no price, dollar amount or free offer appears on the page; no testimonials or customer logos; a "Works with" list of 9 coding agents and 10 platforms, including Poe, NanoGPT, LiteLLM, Dify and models.dev.Source
    • Your websiteWally pricing page (runwally.com, "© 2026 RunAnywhere, Inc."): "Prepaid credits, spent per million tokens." Ten models, from $0.15 input and $0.50 output per million tokens (GLM-5.3 Flash) to $2.925 and $15.00 (Kimi K3), with cached-input prices. The words "free", "welcome", "$5", "trial" and "minimum" do not appear, and there is no FAQ.Source
    • Your websiteDocs: the Wally quickstart has four steps (Install, Sign in, Run against the cloud, Call the API directly) and says only "Cloud usage bills against prepaid credits"; an account out of credit gets a 403 "insufficient_credit". The docs index (llms.txt, 180+ pages) has no pricing, rate-limit or changelog page.Source
  • 02

    Prices live only on runwally.com/pricing. runanywhere.ai/pricing returns 404, the sitemap has no pricing page, and the 180+ page docs index (llms.txt) has no pricing, rate-limit or changelog page. Developers who type the obvious URL, and coding agents that read llms.txt to choose a provider, find no price.

    High impactLow effort First result in 1 week Co founder and CTO

    First action

    Add a redirect from runanywhere.ai/pricing to runwally.com/pricing.

    How to measure it

    Pricing-to-sign-up rate

    Now
    –
    Target
    +20%
    By
    60 days

    Steps

    1. 1

      Redirect

      runanywhere.ai/pricing to runwally.com/pricing.

      Co-founder and CTO 30 minutes

    2. 2

      One price file

      A single source for the price table on both sites and the docs.

      Co-founder and CTO 1 day

    3. 3

      Docs page

      Pricing and limits: prices, cached input, rate limits, running out.

      Co-founder and CTO 1 day

    4. 4

      llms.txt

      List the page with a clear one-line summary.

      Co-founder and CTO 30 minutes

    5. 5

      Link it

      From the quickstart, the API reference and the homepage footer.

      Developer Relations & Technical Content Intern 2 hours

    6. 6

      Check agents

      Ask two coding agents what Wally costs and confirm they find the page.

      Developer Relations & Technical Content Intern 1 hour

    First result
    1 week
    Cost
    $0
    Owner
    Co founder and CTO

    Smallest test

    Ship the redirect today and watch 404s fall in the logs.

    Why it works

    Redirect runanywhere.ai/pricing to runwally.com/pricing (or show the same table). Add a docs page "Pricing and limits" with the per-model table, cached-input prices, the $5 grant, rate limits and the insufficient_credit behaviour, and list it in llms.txt. Generate both from one price file so they never drift apart.

    Pricing is the page developers look for first, and coding agents now pick providers from machine-readable docs. One source of truth also stops prices drifting between three sites.

    Risks and how to handle them

    • Prices change often

      Generate every table from one file and show an "updated" date.

    Evidence

    • Your websiteSitemap (75 URLs, checked November 9, 2026): 43 blog posts, 21 Inference Radar issues (2026-W13 to W33), product pages /neurt, /qhexrt and /sdk, legal pages and /about. There is no pricing page (runanywhere.ai/pricing returns 404), no page per Wally model, and no comparison, customer or careers pages.Source
    • Your websiteDocs: the Wally quickstart has four steps (Install, Sign in, Run against the cloud, Call the API directly) and says only "Cloud usage bills against prepaid credits"; an account out of credit gets a 403 "insufficient_credit". The docs index (llms.txt, 180+ pages) has no pricing, rate-limit or changelog page.Source
    • Your websiteWally pricing page (runwally.com, "© 2026 RunAnywhere, Inc."): "Prepaid credits, spent per million tokens." Ten models, from $0.15 input and $0.50 output per million tokens (GLM-5.3 Flash) to $2.925 and $15.00 (Kimi K3), with cached-input prices. The words "free", "welcome", "$5", "trial" and "minimum" do not appear, and there is no FAQ.Source
  • 03

    The Apple Neural Engine product is called MetalRT in the nav, the sidebar and all four benchmark posts (9 uses) but NeuRT in the URL, the page title and the footer (2 uses). The Hacker News launch used MetalRT. Search results show "NeuRT | RunAnywhere" while every link and post says MetalRT, so people searching either name may miss the page, and the press can't be sure what the product is called.

    Medium impactLow effort First result in 1 month Co founder and CTO

    First action

    Decide the name and write the one-line "formerly" note if needed.

    How to measure it

    Engine-name search impressions

    Now
    –
    Target
    +25%
    By
    60 days

    Steps

    1. 1

      Decide

      MetalRT or NeuRT, and why.

      Co-founder and CEO 1 hour

    2. 2

      Page and URL

      Title, H1, URL and a redirect from the other name.

      Co-founder and CTO 2 hours

    3. 3

      Nav and footer

      The same name everywhere on runanywhere.ai.

      Co-founder and CTO 1 hour

    4. 4

      Old posts

      A "formerly" note on the four benchmark posts.

      Developer Relations & Technical Content Intern 2 hours

    5. 5

      Off-site

      Docs, GitHub README, Hugging Face and npm package descriptions.

      Developer Relations & Technical Content Intern 1 day

    6. 6

      Check

      Search Console for both names after a month.

      Co-founder and CTO monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder and CTO

    Smallest test

    Make the nav and footer match today.

    Why it works

    Choose one name. Then update the page title, URL (with a redirect from the other), nav, footer, docs and the Hugging Face and GitHub mentions. If the rename is deliberate, add one line ("NeuRT, formerly MetalRT") so old benchmark posts and press still connect.

    One name concentrates search, links and word of mouth on one page, and makes the benchmarks easier to cite.

    Risks and how to handle them

    • Losing links to the old name

      Redirect and keep the "formerly" line for a year.

    Evidence

    • Your websiteApple engine page: the URL is /neurt and the page title "NeuRT | RunAnywhere", but the nav, sidebar and all four benchmark write-ups call it MetalRT (9 uses against 2 for NeuRT), and the footer says NeuRT. Benchmarks on M4 Max: 658 tok/s LLM decode, 6.6 ms time to first token, 287 tok/s VLM decode, 101 ms STT. No availability, licence or price; the only path is "Talk to us".Source
    • Hacker NewsThe MetalRT launch post "Fastest LLM decode engine on Apple Silicon. 658 tok/s on M4-max,beats mlx by 19%" got 5 points and 3 comments; one comment asked whether the engine was AI-generated.Source
  • 04

    runanywhere.ai/newsletter calls Inference Radar a "weekly" briefing on 130+ repos, but its archive ends at 2026-W33 (August 13 to 19). The newsletter itself kept going: inference-radar.com, "Powered by RunAnywhere", already shows 2026-W44. So about eleven issues of fresh, citation-backed inference content are missing from the main domain, and a visitor to /newsletter sees a briefing that looks like it stopped in August. The blog index has no signup form either.

    Medium impactLow effort First result in 2 weeks Developer Relations & Technic…

    First action

    Backfill W34 to W44 on runanywhere.ai/newsletter.

    How to measure it

    Sign-ups from the main domain

    Now
    –
    Target
    200 a month
    By
    90 days

    Steps

    1. 1

      Backfill

      Add W34 to W44 to the archive and sitemap.

      Developer Relations & Technical Content Intern 1 day

    2. 2

      Pick a canonical

      One home per issue, with the other site pointing to it.

      Co-founder and CTO 1 hour

    3. 3

      Automate

      Publish to both sites from the same pipeline each week.

      Co-founder and CTO 1 day

    4. 4

      Inline signup

      On /newsletter, the blog index and post endings.

      Co-founder and CTO 2 hours

    5. 5

      Product line

      One relevant Wally or SDK line per issue.

      Developer Relations & Technical Content Intern weekly

    6. 6

      Measure

      Sign-ups by page and clicks to Wally.

      Developer Relations & Technical Content Intern monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Backfill the last three issues and add one signup box on /newsletter.

    Why it works

    Publish every issue to runanywhere.ai/newsletter automatically when it goes out, with a canonical link to one home so the two sites don't compete in search. Add an inline signup box on /newsletter, on the blog index and at the end of every post, and end each issue with one line about Wally or the SDKs.

    Weekly, specific inference content builds search authority for the domain that sells the product, and an archive that looks current shows the team is active.

    Risks and how to handle them

    • Duplicate content across two sites

      Use one canonical URL per issue.

    Evidence

    • Your websiteInference Radar: /newsletter calls it a free "weekly" briefing on "130+ repos" and lists 21 issues, the latest 2026-W33 (August 13 to 19, 2026). "Subscribe" leaves for inference-radar.com, which is "Powered by RunAnywhere" and already shows issue 2026-W44. The blog index has no signup form.Source
    • Your websiteSitemap (75 URLs, checked November 9, 2026): 43 blog posts, 21 Inference Radar issues (2026-W13 to W33), product pages /neurt, /qhexrt and /sdk, legal pages and /about. There is no pricing page (runanywhere.ai/pricing returns 404), no page per Wally model, and no comparison, customer or careers pages.Source
  • 05

    Wally's strongest proof sits on runwally.com: a chart, sourced from Artificial Analysis, showing an open model on Wally scoring 45 on the Intelligence Index against 58 for a closed-model baseline at about 1/23 of its cost per task. The runanywhere.ai homepage, which the GitHub repo, blog, press and YC profile all point to, shows no number for Wally at all: no price, no chart and no testimonial.

    High impactLow effort First result in 2 weeks Co founder and CEO

    First action

    Export the chart from runwally.com as a static image with source and date.

    How to measure it

    Homepage-to-Wally click rate

    Now
    –
    Target
    +25%
    By
    45 days

    Steps

    1. 1

      Export

      A static chart with source and date.

      Co-founder and CTO 2 hours

    2. 2

      Write the line

      One sentence and a caveat on how cost per task is measured.

      Co-founder and CEO 1 hour

    3. 3

      Place it

      In the Wally card on the homepage, beside "Get started".

      Co-founder and CTO 2 hours

    4. 4

      Reuse

      Catalog post, README and social cards.

      Developer Relations & Technical Content Intern 1 day

    5. 5

      Measure

      Click-through to Wally and sign-ups.

      Co-founder and CTO 2 weeks

    6. 6

      Refresh

      Update when models or prices change.

      Developer Relations & Technical Content Intern monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Co founder and CEO

    Smallest test

    Add the one-line claim with a link to the chart this week.

    Why it works

    Add a compact version of the chart to the Wally section of the runanywhere.ai homepage with one line ("About 1/23 of the cost per task"), the source and date, and a "Start with $5" button. Reuse it in the Wally catalog post, the README and social posts. Refresh the numbers whenever the catalogue changes.

    Developers compare cost per finished task, not just per token, and competitors such as Fireworks already lead with "Pay less per finished task". Showing the number where traffic arrives makes the claim RunAnywhere's own.

    Risks and how to handle them

    • Benchmark claims age quickly

      Date the chart and link the source.

    Evidence

    • Your websiteWally homepage (runwally.com, "Wally, product by RunAnywhere"): "Frontier open models, one API key away." A cost-per-task chart sourced from Artificial Analysis shows DeepSeek V4.1 Flash 1017 scoring 45 on the Intelligence Index against 58 for a closed-model baseline, at about 1/23 of its cost per task. Sign-up goes to the console and the CLI signs in with Google or GitHub. No testimonials.Source
    • Your websiteHomepage (checked November 9, 2026): H1 "We make the best open models faster and cheaper to run, for everyone." The hero button "Get started" goes straight to the console; usage is "paid from prepaid credits you buy in the console"; no price, dollar amount or free offer appears on the page; no testimonials or customer logos; a "Works with" list of 9 coding agents and 10 platforms, including Poe, NanoGPT, LiteLLM, Dify and models.dev.Source
    • Competitor researchOpenRouter: "One API for text, image, video, and audio models", hundreds of models, with free model variants rate limited per day. Together AI: "The AI Native Cloud", with serverless, batch and dedicated inference and published customer results. Fireworks: "Pay less per finished task" and "up to 99.6% cache hits".Source
  • 06

    RunAnywhere's biggest audience is the on-device SDK: 10.3k GitHub stars, 413 forks and a 16-post tutorial series across four platforms. None of it leads to the hosted product. The README mentions the wally CLI but doesn't link runwally.com, and Swift Part 1 ends with "Open an issue on GitHub or reach out on X", with no link to Wally or the console. The "best SDK" round-up doesn't mention Wally either.

    High impactLow effort First result in 2 weeks Developer Relations & Technic…

    First action

    Add the Wally block and a runwally.com link to the README.

    How to measure it

    SDK-sourced Wally sign-ups

    Now
    –
    Target
    100 a month
    By
    60 days

    Steps

    1. 1

      Write the block

      Three lines, one command, the $5 start.

      Developer Relations & Technical Content Intern 2 hours

    2. 2

      README

      Add it under the quickstart with a tagged runwally.com link.

      Co-founder and CTO 1 hour

    3. 3

      Tutorials

      Add it to the 16 platform tutorials.

      Developer Relations & Technical Content Intern 1 day

    4. 4

      Round-ups

      Mention the hosted option in the "best" and "how to run locally" posts.

      Developer Relations & Technical Content Intern 1 day

    5. 5

      Tag sources

      UTM per page, read in the console sign-up data.

      Co-founder and CTO 2 hours

    6. 6

      Review

      Which pages send sign-ups; expand the winners.

      Developer Relations & Technical Content Intern monthly

    First result
    2 weeks
    Cost
    $0
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Add the block to the README and the four Part 1 tutorials only.

    Why it works

    Add a short "When the device isn't enough" block to the README and to the end of each tutorial: run the same prompt on a frontier open model on Wally with one key, start with $5, and copy one command (`wally account login`). Tag each link by source so the team can see which tutorials send users.

    Developers who already trust the SDK are the warmest Wally prospects, and every app that runs on a device also has requests too large for it. The block costs nothing and sits on pages that already get traffic.

    Risks and how to handle them

    • Open-source users dislike upsells

      Keep it one short, useful block and say the SDK stays fully offline without a key.

    Builds on

    Evidence

    • Your GitHubrunanywhere-sdks repository: 10.3k stars, 413 forks, 160 open issues and 57 open pull requests; SDKs "currently 0.20.37"; licence "RunAnywhere License (Apache 2.0 based, with additional commercial-use terms)". The README mentions the wally CLI but does not link runwally.com, and says every SDK "runs fully offline without an API key".Source
    • Your websiteTutorials: the sitemap holds a 16-post series (chat, STT, TTS and voice pipeline for Swift, Kotlin, React Native and Flutter). Swift Part 1 is dated January 23, 2026 and pins SDK 0.17.4; its body has no link to Wally or the console and ends "Questions? Open an issue on GitHub or reach out on X."Source
    • Your website"The 7 Best AI SDKs for On-Device Inference in 2026" (February 13, 2026) ranks RunAnywhere first, ahead of TensorFlow Lite, ONNX Runtime Mobile, Cactus, Liquid AI and Nexa AI. It lists weighted criteria but no per-SDK scores and has no note that the publisher is ranking itself. The sitemap has four more "best" or "top" round-ups.Source
  • 07

    "The 7 Best AI SDKs for On-Device Inference in 2026" ranks RunAnywhere first and lists weighted criteria, but publishes no per-SDK scores and has no note that the publisher is ranking itself. The sitemap holds four more "best" or "top" round-ups. Readers, and AI answer engines that increasingly discount self-ranked lists, have no way to check the ranking.

    Medium impactLow effort First result in 1 month Co founder and CEO

    First action

    Add the disclosure line to the SDK round-up.

    How to measure it

    Round-up organic clicks

    Now
    –
    Target
    +20%
    By
    90 days

    Steps

    1. 1

      Disclose

      One line at the top of all five round-ups.

      Co-founder and CEO 1 hour

    2. 2

      Score table

      Per-SDK scores for each weighted criterion.

      Developer Relations & Technical Content Intern 2 days

    3. 3

      Check rivals

      Each rival's current features from their own site.

      Developer Relations & Technical Content Intern 1 day

    4. 4

      Honest picks

      "Best for" lines where a rival wins.

      Co-founder and CEO 2 hours

    5. 5

      Date them

      "Last checked" dates and a quarterly review.

      Developer Relations & Technical Content Intern 1 hour

    6. 6

      Track

      Clicks and citations in AI answers.

      Developer Relations & Technical Content Intern monthly

    First result
    1 month
    Cost
    $0
    Owner
    Co founder and CEO

    Smallest test

    Disclose and add scores to the SDK round-up first.

    Why it works

    Add a one-line disclosure at the top of each round-up ("We make RunAnywhere; here is how we scored everyone, including us"), publish the score table behind the weights, link each rival's own site, and add a "last checked" date. Where a rival is better for a use case, say so.

    A transparent comparison gets cited and shared, while a self-ranking with no disclosure gets discounted. It also protects the brand with the developer audience it depends on.

    Risks and how to handle them

    • Scores invite argument

      Publish the method and correct errors quickly.

    Evidence

    • Your website"The 7 Best AI SDKs for On-Device Inference in 2026" (February 13, 2026) ranks RunAnywhere first, ahead of TensorFlow Lite, ONNX Runtime Mobile, Cactus, Liquid AI and Nexa AI. It lists weighted criteria but no per-SDK scores and has no note that the publisher is ranking itself. The sitemap has four more "best" or "top" round-ups.Source
    • Competitor researchCactus: "Sub-30MB Foundation Models for Mobile" with "Hybrid routing" cloud fallback, 20.8k GitHub stars, 237k monthly Hugging Face downloads and testimonials from hardware partners.Source
  • 08

    Real visitors already pass Core Web Vitals (75th-percentile LCP 1,373 ms, INP 81 ms, CLS 0), so this is not urgent. But the lab mobile score is 57 with 5,234 ms of total blocking time, fonts are 144 KB of a 472 KB page, render-blocking requests cost about 450 ms and 56 KB of JavaScript is unused. For a company selling speed, a slow lab score is an easy thing for developers to screenshot.

    Low impactLow effort First result in 1 week Co founder and CTO

    First action

    Subset the fonts to the characters used.

    How to measure it

    Mobile PageSpeed score

    Now
    –
    Target
    80+
    By
    30 days

    Steps

    1. 1

      Fonts

      Subset, preload and limit weights.

      Co-founder and CTO 2 hours

    2. 2

      Critical CSS

      Inline above-the-fold styles.

      Co-founder and CTO 2 hours

    3. 3

      Defer scripts

      Drop or defer the 56 KB unused script.

      Co-founder and CTO 2 hours

    4. 4

      Contrast

      Fix the failing colour pair.

      Co-founder and CTO 30 minutes

    5. 5

      Re-test

      PageSpeed mobile and desktop.

      Co-founder and CTO 30 minutes

    6. 6

      Watch real users

      Keep Core Web Vitals passing.

      Co-founder and CTO monthly

    First result
    1 week
    Cost
    $0
    Owner
    Co founder and CTO

    Smallest test

    Subset the fonts and re-run PageSpeed.

    Why it works

    Subset and preload one or two font files, inline critical CSS, defer the unused script, and fix the one colour-contrast failure. Re-test on mobile and keep real-visitor numbers passing.

    Small fixes lift the lab score a developer audience might check, and they improve the first impression on slow phones.

    Risks and how to handle them

    • Design changes

      Keep the same typefaces and only subset them.

    Evidence

    • PageSpeedHomepage, November 10, 2026: lab mobile performance 57 (desktop 66) with 5,234 ms total blocking time, but real visitors pass Core Web Vitals (75th-percentile LCP 1,373 ms, INP 81 ms, CLS 0). Suggested fixes: render-blocking requests (about 450 ms) and 56 KB of unused JavaScript; fonts are 144 KB of the 472 KB page; one colour-contrast issue (accessibility 96).Source

This quarter 12

The core plays for the next three months.

  • 09

    The chat, STT, TTS and voice tutorials for Swift, Kotlin, React Native and Flutter date from January 2026. Swift Part 1 pins SDK 0.17.4, while the repo says the SDKs are "currently 0.20.37", and 160 issues are open. Developers following an old tutorial into a changed API often hit errors and give up.

    Medium impactMedium effort First result in 1 month Developer Relations & Technic…

    First action

    Re-run Swift Part 1 on the current SDK and list what broke.

    How to measure it

    Sample app builds from tutorials

    Now
    –
    Target
    +30%
    By
    1 quarter

    Steps

    1. 1

      Audit

      Run all 16 tutorials on the current SDK.

      Developer Relations & Technical Content Intern 1 week

    2. 2

      Update

      Code, screenshots and version pins.

      Developer Relations & Technical Content Intern 3 weeks

    3. 3

      Stamp

      "Tested on" version and date on each post.

      Developer Relations & Technical Content Intern 1 day

    4. 4

      CI check

      Build the samples on every release.

      Co-founder and CTO 1 week

    5. 5

      Issues to posts

      Top five repeated questions become short guides.

      Developer Relations & Technical Content Intern 2 weeks

    6. 6

      Measure

      Tutorial traffic, sample clones and new issues.

      Developer Relations & Technical Content Intern monthly

    First result
    1 month
    Cost
    Intern time
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Update and stamp the four Part 1 posts.

    Why it works

    Re-run each tutorial on the current SDK, update the code and screenshots, add a "Tested on 0.20.x, <date>" stamp, and link to the docs page for each call. Add a CI job that builds the tutorial sample apps against each release, so breakage shows up before readers find it. Turn the most common GitHub issue questions into short follow-ups, as the intern role describes.

    Working tutorials are the main activation path for an open-source SDK, and current versions rank and convert better than stale ones.

    Risks and how to handle them

    • The SDK keeps changing

      Let CI flag broken tutorials automatically.

    Evidence

    • Your websiteTutorials: the sitemap holds a 16-post series (chat, STT, TTS and voice pipeline for Swift, Kotlin, React Native and Flutter). Swift Part 1 is dated January 23, 2026 and pins SDK 0.17.4; its body has no link to Wally or the console and ends "Questions? Open an issue on GitHub or reach out on X."Source
    • Your GitHubrunanywhere-sdks repository: 10.3k stars, 413 forks, 160 open issues and 57 open pull requests; SDKs "currently 0.20.37"; licence "RunAnywhere License (Apache 2.0 based, with additional commercial-use terms)". The README mentions the wally CLI but does not link runwally.com, and says every SDK "runs fully offline without an API key".Source
    • HiringDeveloper Relations & Technical Content Intern (Y Combinator job board, remote, $750 to $1.5K a month): write tutorials, integration guides and benchmark write-ups, make demo videos and sample apps, engage developers on X, LinkedIn, Reddit, Hacker News, Discord and GitHub, and turn common questions into posts or docs. At 30 days, publish an NPU-versus-CPU benchmark post; by 90 days, own a content calendar item end to end.Source
  • 10

    The repo's licence is "Apache 2.0 based, with additional commercial-use terms", but the /sdk page states no licence, commercial terms, price or customers, and the engines' pages offer only "Talk to us". A team that wants to ship the SDK in a commercial app can't tell whether it needs to pay, so it either holds back or never asks.

    High impactMedium effort First result in 1 month Co founder and CEO

    First action

    Write the three-sentence summary of when a commercial licence is needed.

    How to measure it

    Licence enquiries per month

    Now
    –
    Target
    15
    By
    1 quarter

    Steps

    1. 1

      Summarise

      Who is free and who needs a licence, in plain English.

      Co-founder and CEO 1 day

    2. 2

      Legal check

      Confirm the summary matches the licence text.

      Legal counsel 1 week

    3. 3

      Package

      What a commercial licence includes.

      Co-founder and CEO 2 days

    4. 4

      Publish

      On /sdk and in the README.

      Co-founder and CTO 1 day

    5. 5

      Route

      A short form with company size and platforms.

      Co-founder and CTO 1 day

    6. 6

      Track

      Enquiries to deals.

      Co-founder and CEO monthly

    First result
    1 month
    Cost
    Legal review
    Owner
    Co founder and CEO

    Smallest test

    Add the summary to the README and count licence questions.

    Why it works

    Add a plain-English "Using the SDK commercially" section to /sdk and the README: who can use it free, when a commercial licence is needed, what it includes (engines, support, updates) and either a starting price or a short form that routes to sales. Track how many licence questions turn into paid deals.

    Clear terms turn an existing legal requirement into a revenue path, and serious teams prefer a known price to an open question.

    Risks and how to handle them

    • Scaring off hobbyists

      Say first what stays free.

    Evidence

    • Your GitHubrunanywhere-sdks repository: 10.3k stars, 413 forks, 160 open issues and 57 open pull requests; SDKs "currently 0.20.37"; licence "RunAnywhere License (Apache 2.0 based, with additional commercial-use terms)". The README mentions the wally CLI but does not link runwally.com, and says every SDK "runs fully offline without an API key".Source
    • Your websiteSDK page: "Everything above the kernel is open." One C++ core with six bindings (Swift, Kotlin, React Native, Flutter, TypeScript, C++), a live browser demo and app downloads; no licence, commercial terms, price or named customers on the page.Source
  • 11

    The NPU engines have standout numbers (658 tok/s decode on M4 Max; 12,540 tok/s NPU prefill and 36 ms to first token on Hexagon v81), but neither page states availability, licence or price, and the only path is a call booking. A five-person team can't take calls with every curious developer, and developers who want to try first simply leave.

    Medium impactMedium effort First result in 1 month Co founder and CTO

    First action

    Write the five-field evaluation form and a supported-chips table.

    How to measure it

    Evaluation requests per month

    Now
    –
    Target
    40
    By
    1 quarter

    Steps

    1. 1

      Supported table

      Chips, models and modalities live today.

      Co-founder and CTO 1 day

    2. 2

      Form

      Device, chip, model, use case and volume.

      Co-founder and CTO 1 day

    3. 3

      Evaluation kit

      Build or benchmark script plus terms.

      Co-founder and CTO 2 weeks

    4. 4

      Qualify

      Calls only for high-volume or device-maker requests.

      Co-founder and CEO ongoing

    5. 5

      Follow up

      Ask for results after a week.

      Developer Relations & Technical Content Intern weekly

    6. 6

      Review

      Requests by chip and conversion to deals.

      Co-founder and CEO monthly

    First result
    1 month
    Cost
    Engineering time
    Owner
    Co founder and CTO

    Smallest test

    Add the form to the QHexRT page only.

    Why it works

    Add an "Evaluate" form on both pages asking for the device or chip (M-series, A-series, Hexagon v75/v79/v81), the model and the use case. It returns an evaluation build or a benchmark script by email, and only qualified, high-volume requests get a call. Publish which chips and models are supported today.

    Self-serve evaluation collects more leads and better data than a calendar, and sorts the high-value device makers from the hobbyists.

    Risks and how to handle them

    • Leaking unreleased builds

      Ship benchmark scripts first and builds under evaluation terms.

    Builds on

    Evidence

    • Your websiteApple engine page: the URL is /neurt and the page title "NeuRT | RunAnywhere", but the nav, sidebar and all four benchmark write-ups call it MetalRT (9 uses against 2 for NeuRT), and the footer says NeuRT. Benchmarks on M4 Max: 658 tok/s LLM decode, 6.6 ms time to first token, 287 tok/s VLM decode, 101 ms STT. No availability, licence or price; the only path is "Talk to us".Source
    • Your websiteQHexRT page: "Every modality, 100% on the Hexagon NPU." 12,540 tok/s NPU prefill and 36 ms time to first token (LFM 2.5 230M on Hexagon v81); models pinned per v75, v79 and v81; LLM live, other modalities in development; no pricing or licensing; access through "Talk to us" and Hugging Face.Source
    • Y Combinator profileRunAnywhere, Winter 2026: "Fastest inference anywhere: open models on-prem, hosted or on-device"; team size 5; San Francisco; one open job.Source
  • 12

    The DevRel intern's 30-day goal is a blog post comparing NPU and CPU performance on their own device. The MetalRT launch on Hacker News got 5 points, and one of its 3 comments asked whether the engine was AI-generated. Benchmark claims from a vendor get questioned. One that readers can rerun on their own phone gets shared.

    Medium impactMedium effort First result in 6 weeks Developer Relations & Technic…

    First action

    Turn the intern's benchmark into a script with a fixed model and prompt.

    How to measure it

    Benchmark submissions

    Now
    –
    Target
    300
    By
    1 quarter

    Steps

    1. 1

      Fix the test

      Model, prompt, settings and output format.

      Co-founder and CTO 2 days

    2. 2

      Script

      One command on Mac, Android and iOS.

      Developer Relations & Technical Content Intern 2 weeks

    3. 3

      Board

      A public page of results by device.

      Co-founder and CTO 1 week

    4. 4

      Seed

      The team's own devices first.

      Developer Relations & Technical Content Intern 2 days

    5. 5

      Launch

      Post with the write-up on Hacker News, Reddit and X.

      Developer Relations & Technical Content Intern 1 day

    6. 6

      Feed back

      Monthly highlights in Inference Radar and the engine pages.

      Developer Relations & Technical Content Intern monthly

    First result
    6 weeks
    Cost
    Intern and engineering time
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Publish the script with the post and count reruns.

    Why it works

    Package the benchmark as a one-command script and a browser or app page that runs the same model on CPU and NPU and posts the result to a public board (device, chip, tok/s, time to first token). Seed it with the team's devices, invite the community on Discord, Reddit and X, and update the engine pages from the board.

    Community results are more credible than vendor charts, create a steady stream of posts, and collect device data the engines team can use.

    Risks and how to handle them

    • Bad or fake submissions

      Accept signed results from the script only and flag outliers.

    Evidence

    • HiringDeveloper Relations & Technical Content Intern (Y Combinator job board, remote, $750 to $1.5K a month): write tutorials, integration guides and benchmark write-ups, make demo videos and sample apps, engage developers on X, LinkedIn, Reddit, Hacker News, Discord and GitHub, and turn common questions into posts or docs. At 30 days, publish an NPU-versus-CPU benchmark post; by 90 days, own a content calendar item end to end.Source
    • Hacker NewsThe MetalRT launch post "Fastest LLM decode engine on Apple Silicon. 658 tok/s on M4-max,beats mlx by 19%" got 5 points and 3 comments; one comment asked whether the engine was AI-generated.Source
    • Your websiteQHexRT page: "Every modality, 100% on the Hexagon NPU." 12,540 tok/s NPU prefill and 36 ms time to first token (LFM 2.5 230M on Hexagon v81); models pinned per v75, v79 and v81; LLM live, other modalities in development; no pricing or licensing; access through "Talk to us" and Hugging Face.Source
  • 13

    Wally serves ten models, each with a fast variant under its own id, but there is no page per model anywhere: the sitemap has none, the docs index lists no models, and prices sit in one table on runwally.com. Developers search by model name plus "API", "price" or "provider", and coding agents look for a model id and base URL. Today those searches land on aggregators.

    High impactMedium effort First result in 6 weeks Developer Relations & Technic…

    First action

    Build the page for GLM-5.3 Flash, the cheapest model, as the template.

    How to measure it

    Model-page sign-ups

    Now
    –
    Target
    150 a month
    By
    1 quarter

    Steps

    1. 1

      Template

      Price, speed, context, privacy, setup.

      Developer Relations & Technical Content Intern 2 days

    2. 2

      Measure speed

      Tokens per second and time to first token per model.

      Co-founder and CTO 3 days

    3. 3

      Generate

      Ten pages from the shared data file.

      Co-founder and CTO 1 week

    4. 4

      Setup snippets

      CLI, API and two coding agents per model.

      Developer Relations & Technical Content Intern 3 days

    5. 5

      Link

      From the catalog post, docs, sitemap and llms.txt.

      Developer Relations & Technical Content Intern 1 day

    6. 6

      Track

      Rankings and sign-ups per page.

      Developer Relations & Technical Content Intern monthly

    First result
    6 weeks
    Cost
    Staff time only
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Publish one model page and compare its sign-ups with the pricing table's.

    Why it works

    Generate one page per model on runwally.com (or runanywhere.ai/models): price per million tokens, cached input, the fast variant, measured tokens per second and time to first token, context length, zero data retention, and copy-paste setup for the CLI, the API and two coding agents. Build them from the same price file as idea 2 and link them from the catalog post and the docs.

    Model-name searches are the highest-intent traffic in inference, and pages built from one data file stay accurate as the catalogue changes.

    Risks and how to handle them

    • Speed numbers vary with load

      Show a dated median and how it was measured.

    Builds on

    Evidence

    • Your website"Wally now serves ten open models" (October 22, 2026): each model gets a fast version, "the same model, served for speed", under its own id, plus Eve, a decision model run with `wally decisions`. The post sends readers to runwally.com/pricing for rates and to the console to buy prepaid credits.Source
    • Your websiteSitemap (75 URLs, checked November 9, 2026): 43 blog posts, 21 Inference Radar issues (2026-W13 to W33), product pages /neurt, /qhexrt and /sdk, legal pages and /about. There is no pricing page (runanywhere.ai/pricing returns 404), no page per Wally model, and no comparison, customer or careers pages.Source
    • Competitor researchOpenRouter: "One API for text, image, video, and audio models", hundreds of models, with free model variants rate limited per day. Together AI: "The AI Native Cloud", with serverless, batch and dedicated inference and published customer results. Fireworks: "Pay less per finished task" and "up to 99.6% cache hits".Source
  • 14

    Wally is pitched for coding agents and "fast versions for agent loops", and it bills from prepaid credit. When the balance hits zero, requests fail with a 403 "insufficient_credit", often in the middle of an unattended agent run. That failure is the moment a developer is most likely to switch providers, and it happens right after they have shown they are a heavy user.

    High impactMedium effort First result in 1 month Co founder and CTO

    First action

    Count accounts that hit insufficient_credit last month and how many topped up after.

    How to measure it

    Top-up rate after a low balance

    Now
    –
    Target
    60%
    By
    1 quarter

    Steps

    1. 1

      Measure

      insufficient_credit events and what happened next.

      Co-founder and CTO 1 day

    2. 2

      Alerts

      Email and CLI warnings at two thresholds.

      Co-founder and CTO 1 week

    3. 3

      Auto top-up

      Opt-in with a monthly cap through Stripe.

      Co-founder and CTO 1 week

    4. 4

      Grace buffer

      Let the running request finish.

      Co-founder and CTO 3 days

    5. 5

      CLI summary

      Balance and spend after each session.

      Co-founder and CTO 2 days

    6. 6

      Review

      Churn after zero balance, monthly.

      Co-founder and CEO monthly

    First result
    1 month
    Cost
    Engineering time
    Owner
    Co founder and CTO

    Smallest test

    Send a low-balance email at 20% for one month.

    Why it works

    Add low-balance alerts (email and a CLI warning at 20% and 5% of the last top-up), an optional auto top-up with a monthly cap, and a short grace buffer so a running task can finish. Show the balance and spend rate after every `wally` session. Track how many accounts hit insufficient_credit and how many come back.

    Prepaid billing only works if the money never runs out by surprise. Auto top-up also turns one-off buyers into recurring revenue.

    Risks and how to handle them

    • Surprise charges

      Opt-in only, with a cap and an email receipt each time.

    Evidence

    • Your websiteDocs: the Wally quickstart has four steps (Install, Sign in, Run against the cloud, Call the API directly) and says only "Cloud usage bills against prepaid credits"; an account out of credit gets a 403 "insufficient_credit". The docs index (llms.txt, 180+ pages) has no pricing, rate-limit or changelog page.Source
    • Your website"Wally now serves ten open models" (October 22, 2026): each model gets a fast version, "the same model, served for speed", under its own id, plus Eve, a decision model run with `wally decisions`. The post sends readers to runwally.com/pricing for rates and to the console to buy prepaid credits.Source
    • Your websiteRefunds policy: "The $5 welcome grant is promotional credit, not a purchase, and is not refundable." Purchased credit "does not expire while your account is open"; prepaid Wally credits bought through Stripe are non-refundable except for billing errors, duplicate charges and platform faults.Source
  • 15

    Wally sells one thing: prepaid credit spent per million tokens, with no plans, packs or minimums on the pricing page. Purchased credit doesn't expire while the account is open, which makes larger prepaid packs low-risk for buyers. Teams running several agents today must each buy and watch their own balance.

    Medium impactMedium effort First result in 1 month Co founder and CEO

    First action

    Look at current top-up sizes and set the pack levels just above the most common one.

    How to measure it

    Average prepaid purchase

    Now
    –
    Target
    +40%
    By
    1 quarter

    Steps

    1. 1

      Study top-ups

      Current sizes and frequency.

      Co-founder and CEO 1 day

    2. 2

      Set packs

      Three sizes and bonus levels.

      Co-founder and CEO 1 day

    3. 3

      Build checkout

      Packs in the console through Stripe.

      Co-founder and CTO 1 week

    4. 4

      Team accounts

      Shared balance, per-key limits, export.

      Co-founder and CTO 3 weeks

    5. 5

      Show it

      Packs on the pricing page.

      Co-founder and CTO 1 day

    6. 6

      Measure

      Purchase size and repeat buys.

      Co-founder and CEO monthly

    First result
    1 month
    Cost
    Bonus credit cost and engineering time
    Owner
    Co founder and CEO

    Smallest test

    Offer one $100 pack with 5% bonus for a month.

    Why it works

    Add three packs (for example $20, $100 and $500, with 0%, 5% and 10% bonus credit) and a team account with one balance, per-key spend limits and a usage export. Show the packs on the pricing page next to the $5 starting credit.

    Bonus packs pull revenue forward and raise average purchase size, and shared team balances move Wally from one developer's card to a company budget.

    Risks and how to handle them

    • Bonuses eat margin

      Keep bonuses below the gross margin on the cheapest model.

    Builds on

    Evidence

    • Your websiteWally pricing page (runwally.com, "© 2026 RunAnywhere, Inc."): "Prepaid credits, spent per million tokens." Ten models, from $0.15 input and $0.50 output per million tokens (GLM-5.3 Flash) to $2.925 and $15.00 (Kimi K3), with cached-input prices. The words "free", "welcome", "$5", "trial" and "minimum" do not appear, and there is no FAQ.Source
    • Your websiteRefunds policy: "The $5 welcome grant is promotional credit, not a purchase, and is not refundable." Purchased credit "does not expire while your account is open"; prepaid Wally credits bought through Stripe are non-refundable except for billing errors, duplicate charges and platform faults.Source
  • 16

    Zero data retention is a headline feature ("Wally does not store prompts or replies"), the footer shows a SOC 2 Type II audit in progress, and the press list includes "Regulated Industries Are Rewriting Their AI Architecture". But the facts a security reviewer needs are spread across the homepage, docs and legal pages.

    Medium impactMedium effort First result in 1 month Co founder and CEO

    First action

    List the ten questions security reviewers ask most.

    How to measure it

    Company-domain sign-ups

    Now
    –
    Target
    +30%
    By
    1 quarter

    Steps

    1. 1

      Collect questions

      From calls and support email.

      Co-founder and CEO 2 days

    2. 2

      Write

      Retention, logging, regions, sub-processors.

      Co-founder and CTO 3 days

    3. 3

      Legal check

      Match the privacy policy and terms.

      Legal counsel 1 week

    4. 4

      Publish

      Linked from Wally, pricing and docs.

      Co-founder and CTO 1 day

    5. 5

      Update SOC 2

      Swap the status when the report is issued.

      Co-founder and CEO as needed

    6. 6

      Use in sales

      Send before every company call.

      Co-founder and CEO ongoing

    First result
    1 month
    Cost
    Legal review
    Owner
    Co founder and CEO

    Smallest test

    Publish a short FAQ version in the docs first.

    Why it works

    Write one page: what is and isn't stored, logging and metadata, where inference runs, sub-processors, the SOC 2 status and its trust-centre link, and how to request a security review. Link it from the Wally hero, the pricing page and the docs. Add it to sales emails for larger accounts.

    Teams handling sensitive code or data buy on trust as well as price, and one clear page shortens their review.

    Risks and how to handle them

    • Overclaiming

      State only what is in place today and date the page.

    Evidence

    • Your websiteTrust details: the homepage says Wally has "zero data retention by default" and does not store prompts or replies; the footer shows a SOC 2 Type II badge marked "Audit in progress"; the About page press list includes "Regulated Industries Are Rewriting Their AI Architecture" (NextTech Today, March 9, 2026).Source
    • Your websiteAbout page: founded in 2025 "to run AI on the devices people already own"; Y Combinator Winter 2026; engines built March to June 2026; Wally launched in September 2026 and grew to ten models in October; "a small team of engineers who direct AI agents". The press list runs from January to March 2026 and is all about on-device AI.Source
  • 17

    Neither runanywhere.ai nor runwally.com shows a testimonial or customer logo. The published stories (LLM Hub's iOS port in about six and a half weeks, and PickleRite) are both about the on-device SDK, not Wally. Competitors such as Together AI lead with customer results, and Cactus shows partner testimonials.

    High impactMedium effort First result in 6 weeks Co founder and CEO

    First action

    Pull the top 20 Wally accounts by 30-day spend and pick five to ask.

    How to measure it

    Wally page sign-up rate

    Now
    –
    Target
    +15%
    By
    1 quarter

    Steps

    1. 1

      Find users

      Top accounts by steady spend.

      Co-founder and CEO 1 day

    2. 2

      Ask

      Credit in return for a story and numbers.

      Co-founder and CEO 1 week

    3. 3

      Interview

      Cost per task, speed, setup time.

      Developer Relations & Technical Content Intern 2 weeks

    4. 4

      Write

      Short posts with the numbers.

      Developer Relations & Technical Content Intern 2 weeks

    5. 5

      Place

      Quotes on the Wally hero and pricing.

      Co-founder and CTO 1 day

    6. 6

      Share

      X, LinkedIn and Inference Radar.

      Developer Relations & Technical Content Intern 1 week

    First result
    6 weeks
    Cost
    Credit incentives
    Owner
    Co founder and CEO

    Smallest test

    One quote with one number on the pricing page.

    Why it works

    Use console data to find three accounts with steady Wally usage (a coding-agent user, an app on the API and a platform partner). Offer credit in return for a short story with before-and-after cost per task, speed and setup time. Publish them as posts and as one-line quotes on the Wally hero and pricing page.

    Developers believe other developers' numbers more than vendor charts, and the stories also show what each model is good for.

    Risks and how to handle them

    • Users don't want to be named

      Offer anonymous stories with numbers and an industry label.

    Evidence

    • Your websitePublished case study "Android to iOS in Six Weeks, with the RunAnywhere SDK" (July 14, 2026): LLM Hub, a solo developer's on-device app earning about $2K a month on Android, shipped on iOS about six and a half weeks after the first commit, with six features on one SDK. The other published story is PickleRite (March 13, 2026). Neither is about Wally.Source
    • Your websiteWally homepage (runwally.com, "Wally, product by RunAnywhere"): "Frontier open models, one API key away." A cost-per-task chart sourced from Artificial Analysis shows DeepSeek V4.1 Flash 1017 scoring 45 on the Intelligence Index against 58 for a closed-model baseline, at about 1/23 of its cost per task. Sign-up goes to the console and the CLI signs in with Google or GitHub. No testimonials.Source
    • Competitor researchOpenRouter: "One API for text, image, video, and audio models", hundreds of models, with free model variants rate limited per day. Together AI: "The AI Native Cloud", with serverless, batch and dedicated inference and published customer results. Fireworks: "Pay less per finished task" and "up to 99.6% cache hits".Source
  • 18

    The homepage lists 9 coding agents and 10 platforms (including Poe, NanoGPT, LiteLLM, Dify and models.dev). Some of these send traffic to Wally's own console and some resell Wally inside their own product. With a five-person team, effort spent on every integration equally is effort wasted on most of them.

    Medium impactLow effort First result in 45 days Co founder and CTO

    First action

    Add a source tag to each integration's docs setup snippet.

    How to measure it

    Attributed integration spend

    Now
    –
    Target
    Top three identified
    By
    45 days

    Steps

    1. 1

      Tag links

      UTM on every docs integration link.

      Developer Relations & Technical Content Intern 1 day

    2. 2

      Read user-agents

      Attribute API traffic by calling tool.

      Co-founder and CTO 2 days

    3. 3

      Dashboard

      Sign-ups, requests and spend per source.

      Co-founder and CTO 2 days

    4. 4

      Rank

      After 30 days.

      Co-founder and CEO 1 day

    5. 5

      Co-market

      A joint post or video with the top three.

      Developer Relations & Technical Content Intern 1 month

    6. 6

      Prune

      Basic docs only for the rest.

      Co-founder and CTO 1 day

    First result
    45 days
    Cost
    $0
    Owner
    Co founder and CTO

    Smallest test

    Tag the five most-used setup snippets this week.

    Why it works

    Tag every integration's setup link and API user-agent so the console can attribute sign-ups, requests and spend to each source. After 30 days, rank them, then co-market with the top three (a joint post, a featured listing or a setup video) and stop maintaining the bottom ones beyond basic docs.

    Distribution through other tools is the cheapest acquisition a small inference company has, but only if the team knows which tools convert.

    Risks and how to handle them

    • Resold traffic is hard to see

      Ask platform partners for monthly usage reports.

    Evidence

    • Your websiteHomepage (checked November 9, 2026): H1 "We make the best open models faster and cheaper to run, for everyone." The hero button "Get started" goes straight to the console; usage is "paid from prepaid credits you buy in the console"; no price, dollar amount or free offer appears on the page; no testimonials or customer logos; a "Works with" list of 9 coding agents and 10 platforms, including Poe, NanoGPT, LiteLLM, Dify and models.dev.Source
    • Y Combinator profileRunAnywhere, Winter 2026: "Fastest inference anywhere: open models on-prem, hosted or on-device"; team size 5; San Francisco; one open job.Source
  • 19

    Wally already gives every new account $5 of credit, and its users are developers who share tools on X, Discord and GitHub. There is no referral programme: the console is for buying credits and managing keys, and nothing rewards a user for bringing a teammate.

    Medium impactMedium effort First result in 1 month Co founder and CTO

    First action

    Decide the trigger (first $10 purchase) and the reward.

    How to measure it

    Referral share of new payers

    Now
    –
    Target
    15%
    By
    1 quarter

    Steps

    1. 1

      Rules

      Trigger, reward and limits.

      Co-founder and CEO 1 day

    2. 2

      Links

      Referral codes in the console.

      Co-founder and CTO 1 week

    3. 3

      CLI

      `wally invite` and balance line.

      Co-founder and CTO 3 days

    4. 4

      Fraud checks

      One reward per payment method.

      Co-founder and CTO 2 days

    5. 5

      Announce

      Discord, X and Inference Radar.

      Developer Relations & Technical Content Intern 1 day

    6. 6

      Measure

      Referred payers and cost per payer.

      Co-founder and CEO monthly

    First result
    1 month
    Cost
    Reward credit
    Owner
    Co founder and CTO

    Smallest test

    Give Discord members referral links by hand for a month.

    Why it works

    Give each account a referral link and a `wally invite` command. The new user gets the usual $5 plus $5 more after their first $10 purchase, and the referrer gets $5 at the same moment. Show referral credit in the CLI balance line.

    Paying the reward only after a purchase keeps it cheap and fraud-resistant, and developer tools spread best from one developer to another.

    Risks and how to handle them

    • Self-referral

      Reward only after a purchase from a different payment method.

    Builds on

    Evidence

    • Your websiteRefunds policy: "The $5 welcome grant is promotional credit, not a purchase, and is not refundable." Purchased credit "does not expire while your account is open"; prepaid Wally credits bought through Stripe are non-refundable except for billing errors, duplicate charges and platform faults.Source
    • HiringDeveloper Relations & Technical Content Intern (Y Combinator job board, remote, $750 to $1.5K a month): write tutorials, integration guides and benchmark write-ups, make demo videos and sample apps, engage developers on X, LinkedIn, Reddit, Hacker News, Discord and GitHub, and turn common questions into posts or docs. At 30 days, publish an NPU-versus-CPU benchmark post; by 90 days, own a content calendar item end to end.Source
  • 20

    Every press item listed on the About page is from January to March 2026 and about on-device AI. Wally, launched in September, has no coverage listed. The engineering posts are strong (serving GLM-5.3-Flash on one Trainium2 box, NPU engines), but the main Hacker News launch got 5 points. The content exists; the distribution doesn't.

    Medium impactMedium effort First result in 1 month Developer Relations & Technic…

    First action

    Write a launch kit for the Trainium2 post.

    How to measure it

    Launch referral visits

    Now
    –
    Target
    5,000 a month
    By
    1 quarter

    Steps

    1. 1

      Kit template

      Summary, number, snippet, chart, posts.

      Developer Relations & Technical Content Intern 2 days

    2. 2

      First launch

      The Trainium2 post.

      Developer Relations & Technical Content Intern 1 week

    3. 3

      Answer live

      The CTO in the comments for the first hours.

      Co-founder and CTO per launch

    4. 4

      Pitch

      Five infra and developer newsletters on Wally.

      Co-founder and CEO 2 weeks

    5. 5

      Press list

      Add Wally coverage to the About page.

      Developer Relations & Technical Content Intern ongoing

    6. 6

      Rhythm

      One launch every two weeks.

      Developer Relations & Technical Content Intern ongoing

    First result
    1 month
    Cost
    Staff time only
    Owner
    Developer Relations & Technical Content Intern

    Smallest test

    Relaunch one post with a full kit and compare traffic.

    Why it works

    For each technical post, prepare a launch kit: a short summary with the headline number, a reproducible snippet, a chart, and posts for Hacker News, Reddit, X and LinkedIn, sent at a good time with the team ready to answer questions. Pitch the Wally story (open models at a fraction of the cost per task) to infra and developer newsletters, and add a Wally section to the press list.

    Technical audiences reward reproducible numbers and fast, honest answers, and a regular launch rhythm compounds followers for the next one.

    Risks and how to handle them

    • Seen as spammy

      Launch only posts with a new, reproducible result.

    Evidence

    • Your websiteAbout page: founded in 2025 "to run AI on the devices people already own"; Y Combinator Winter 2026; engines built March to June 2026; Wally launched in September 2026 and grew to ten models in October; "a small team of engineers who direct AI agents". The press list runs from January to March 2026 and is all about on-device AI.Source
    • Hacker NewsThe MetalRT launch post "Fastest LLM decode engine on Apple Silicon. 658 tok/s on M4-max,beats mlx by 19%" got 5 points and 3 comments; one comment asked whether the engine was AI-generated.Source
    • HiringDeveloper Relations & Technical Content Intern (Y Combinator job board, remote, $750 to $1.5K a month): write tutorials, integration guides and benchmark write-ups, make demo videos and sample apps, engage developers on X, LinkedIn, Reddit, Hacker News, Discord and GitHub, and turn common questions into posts or docs. At 30 days, publish an NPU-versus-CPU benchmark post; by 90 days, own a content calendar item end to end.Source

Big bets 3

Larger plays that can change the trajectory.

  • 21

    RunAnywhere's YC line is "open models on-prem, hosted or on-device", but today the SDK and Wally are separate products: the SDK page says cloud and on-device modes are chosen explicitly and cloud keys aren't device credentials, and the README stresses that the SDKs run offline without a key. Cactus already sells "Hybrid routing" with cloud fallback. Only RunAnywhere owns both its own device engines and a frontier hosted catalogue.

    High impactHigh effort First result in 1 quarter Co founder and CTO

    First action

    Prototype fallback for chat in the Swift SDK with GLM-5.3 Flash on Wally.

    How to measure it

    Apps using hybrid routing

    Now
    –
    Target
    100
    By
    2 quarters

    Steps

    1. 1

      Design

      Routing rules, privacy defaults and keys.

      Co-founder and CTO 2 weeks

    2. 2

      Prototype

      Chat fallback in one SDK.

      Co-founder and CTO 1 month

    3. 3

      Pilot

      Five SDK users, including a published one.

      Co-founder and CEO 1 month

    4. 4

      Measure

      Cost, latency and fallback rate.

      Co-founder and CTO ongoing

    5. 5

      Roll out

      All six bindings.

      Co-founder and CTO 1 quarter

    6. 6

      Launch

      Post, docs and benchmark.

      Developer Relations & Technical Content Intern 2 weeks

    First result
    1 quarter
    Cost
    One to two engineers for a quarter
    Owner
    Co founder and CTO

    Smallest test

    A sample app with manual fallback to Wally.

    Why it works

    Add a routing option to the SDK: run on the device when the model and hardware fit, and send the request to a matching Wally model when the prompt is too long, the device too weak or the task too hard. One key, one bill, with clear privacy rules (device-only by default, fallback opt-in). Publish the routing policy and measured cost and latency for both paths.

    Every SDK user becomes a potential Wally customer, and Wally gains a feature pure-cloud rivals can't match. It is also the product the YC tagline already promises.

    Risks and how to handle them

    • Breaks the offline promise

      Keep device-only the default and require explicit opt-in.

    • Spreads a small team thin

      Ship one modality and one SDK first.

    Builds on

    Evidence

    • Y Combinator profileRunAnywhere, Winter 2026: "Fastest inference anywhere: open models on-prem, hosted or on-device"; team size 5; San Francisco; one open job.Source
    • Your GitHubrunanywhere-sdks repository: 10.3k stars, 413 forks, 160 open issues and 57 open pull requests; SDKs "currently 0.20.37"; licence "RunAnywhere License (Apache 2.0 based, with additional commercial-use terms)". The README mentions the wally CLI but does not link runwally.com, and says every SDK "runs fully offline without an API key".Source
    • Competitor researchCactus: "Sub-30MB Foundation Models for Mobile" with "Hybrid routing" cloud fallback, 20.8k GitHub stars, 237k monthly Hugging Face downloads and testimonials from hardware partners.Source
  • 22

    RunAnywhere just published how it serves GLM-5.3-Flash on a single Trainium2 box with its own NKI megakernels, and its YC line promises "on-prem" as well as hosted and on-device. Teams with steady agent workloads or strict data rules often want reserved capacity or their own deployment, and those are the largest contracts in inference.

    High impactHigh effort First result in 1 quarter Co founder and CEO

    First action

    Price one dedicated GLM-5.3-Flash box per month, including margin.

    How to measure it

    Dedicated deployments

    Now
    –
    Target
    3
    By
    2 quarters

    Steps

    1. 1

      Price

      Cost and margin for one box per model.

      Co-founder and CEO 1 week

    2. 2

      Package

      SLA, support and setup.

      Co-founder and CTO 2 weeks

    3. 3

      Page

      A "Dedicated" section on the Wally site.

      Co-founder and CTO 1 week

    4. 4

      Target

      Top prepaid accounts by spend.

      Co-founder and CEO 2 weeks

    5. 5

      Pilot

      One customer for 30 days.

      Co-founder and CTO 1 month

    6. 6

      Case study

      Publish results if permitted.

      Developer Relations & Technical Content Intern 1 month

    First result
    1 quarter
    Cost
    Reserved hardware for the pilot
    Owner
    Co founder and CEO

    Smallest test

    Offer reserved capacity to the top three accounts by email.

    Why it works

    Package a "dedicated Wally" offer: one model on reserved hardware in RunAnywhere's cloud or the customer's account, with a monthly price, an SLA and the same API and CLI. Start with the model and box already proven in the Trainium2 post. Sell it to the heaviest prepaid accounts and to procurement-led teams found through idea 16.

    It turns kernel work into contract revenue, and moves heavy users from prepaid top-ups to predictable monthly bills.

    Risks and how to handle them

    • Operations load on a small team

      Offer one model and one region at first.

    Builds on

    Evidence

    • Y Combinator profileRunAnywhere, Winter 2026: "Fastest inference anywhere: open models on-prem, hosted or on-device"; team size 5; San Francisco; one open job.Source
    • Your websiteAbout page: founded in 2025 "to run AI on the devices people already own"; Y Combinator Winter 2026; engines built March to June 2026; Wally launched in September 2026 and grew to ten models in October; "a small team of engineers who direct AI agents". The press list runs from January to March 2026 and is all about on-device AI.Source
    • Competitor researchOpenRouter: "One API for text, image, video, and audio models", hundreds of models, with free model variants rate limited per day. Together AI: "The AI Native Cloud", with serverless, batch and dedicated inference and published customer results. Fireworks: "Pay less per finished task" and "up to 99.6% cache hits".Source
  • 23

    The homepage quotes the CEO, "Access to intelligence is a fundamental human right", and the October post argues that cheap tokens don't matter without usable access and that "neither wealth nor popularity should shut out a legitimate new founder". The post proposes ideas but launches nothing for founders. A mission this public is also a growth channel, if there is a programme behind it.

    High impactMedium effort First result in 1 quarter Co founder and CEO

    First action

    Set a monthly grant budget in credit and write a five-question application.

    How to measure it

    Grant-to-paid conversion

    Now
    –
    Target
    25%
    By
    2 quarters

    Steps

    1. 1

      Budget

      Monthly credit pool at serving cost.

      Co-founder and CEO 1 day

    2. 2

      Criteria

      Simple, public and not based on wealth or followers.

      Co-founder and CEO 2 days

    3. 3

      Application

      Five questions and a monthly review.

      Developer Relations & Technical Content Intern 2 days

    4. 4

      Announce

      A post that links back to the access essay.

      Co-founder and CEO 1 week

    5. 5

      Report

      Monthly totals and what was built.

      Developer Relations & Technical Content Intern monthly

    6. 6

      Convert

      Offer packs when grants run out.

      Co-founder and CEO ongoing

    First result
    1 quarter
    Cost
    Credit at serving cost
    Owner
    Co founder and CEO

    Smallest test

    Ten grants in one month, announced on Discord.

    Why it works

    Launch a small, public grant programme: for example, $100 to $500 of Wally credit each month for early founders, open-source maintainers and student builders, chosen by a short application and with no wealth or follower test. Publish who it serves (by category, not by name), how much has been granted and what was built. Make recipients the first people asked for stories and feedback.

    It gives the mission substance, brings in builders before they choose a provider, and creates stories and goodwill that paid ads can't buy. The cost is credit at RunAnywhere's own serving cost.

    Risks and how to handle them

    • Abuse of free credit

      Human review, monthly caps and the existing acceptable-use policy.

    Evidence

    • Your website"Cheap Tokens Don't Matter Without Access" (October 27, 2026) argues that startups and defenders need usable access, not just cheap tokens, and that "neither wealth nor popularity should shut out a legitimate new founder"; it opens a KYC-reviewed request path on Wally for authorised security teams. The homepage quotes the CEO: "Access to intelligence is a fundamental human right."Source
    • Your websiteRefunds policy: "The $5 welcome grant is promotional credit, not a purchase, and is not refundable." Purchased credit "does not expire while your account is open"; prepaid Wally credits bought through Stripe are non-refundable except for billing errors, duplicate charges and platform faults.Source

Wildcards 1

Unexpected experiments, each with a cheap way to test it.

  • 24

    Wally's catalogue includes Eve, a decision model for yes/no, choice and score questions through `wally decisions` and POST /v1/decisions, priced at $0.042 per million input tokens. It is the cheapest and most unusual thing Wally sells, and most developers have never seen a model built only to decide. It is mentioned once in the catalog post and the docs.

    Medium impactMedium effort First result in 1 month Co founder and CTO

    First action

    Build the GitHub Action version that scores pull request descriptions.

    How to measure it

    Eve-sourced sign-ups

    Now
    –
    Target
    300
    By
    1 quarter

    Steps

    1. 1

      Pick uses

      Three everyday decisions developers make.

      Developer Relations & Technical Content Intern 1 day

    2. 2

      Build page

      Question, options, decision, cost.

      Co-founder and CTO 1 week

    3. 3

      GitHub Action

      Score PR descriptions with Eve.

      Co-founder and CTO 1 week

    4. 4

      Rate limit

      Anonymous cap, then sign-up.

      Co-founder and CTO 2 days

    5. 5

      Launch

      Hacker News, X and Discord.

      Developer Relations & Technical Content Intern 1 day

    6. 6

      Measure

      Uses, shares and sign-ups.

      Developer Relations & Technical Content Intern monthly

    First result
    1 month
    Cost
    Low: Eve calls cost fractions of a cent
    Owner
    Co founder and CTO

    Smallest test

    Ship the GitHub Action and post it once.

    Why it works

    Build a small public page (and a GitHub Action) where anyone can paste a question with options and get Eve's decision and score: for example "Is this pull request description clear enough?" or "Which of these three titles is best?". Show the cost of each call in fractions of a cent and a one-line API snippet. Rate-limit it and point heavier users to a free account with $5.

    A free, shareable tool with an odd, useful output spreads on developer social media, and every use shows Wally's price and API.

    Risks and how to handle them

    • Misused for decisions about people

      Block such uses in the tool and point to the acceptable-use policy.

    Builds on

    Evidence

    • Your website"Wally now serves ten open models" (October 22, 2026): each model gets a fast version, "the same model, served for speed", under its own id, plus Eve, a decision model run with `wally decisions`. The post sends readers to runwally.com/pricing for rates and to the console to buy prepaid credits.Source
    • Your websiteWally pricing page (runwally.com, "© 2026 RunAnywhere, Inc."): "Prepaid credits, spent per million tokens." Ten models, from $0.15 input and $0.50 output per million tokens (GLM-5.3 Flash) to $2.925 and $15.00 (Kimi K3), with cached-input prices. The words "free", "welcome", "$5", "trial" and "minimum" do not appear, and there is no FAQ.Source

03

How Programmable can help

Where an idea in this playbook is something Programmable builds or runs, it's listed here with what we'd do. Everything else is yours to run.

ProgrammableInfo

Turn the research into a list you can act on

These ideas start with finding the right companies. Info finds them and the right person at each, researches every one and drafts the outreach, then asks before it sends.

  • Build this list in Info for idea #15 (opens programmable.info in a new tab)

New accounts: 100,000 free credits with code BETA100K

Find the companies (opens programmable.info in a new tab)
More example playbooks and reviews
ProgrammableDigital

Show RunAnywhere’s product working

Turn screenshots or the website into an animated product demo in its own brand, for the homepage, launch posts and outreach in this playbook.

Make a demo (opens programmable.digital in a new tab)

More growth playbooks

Other companies we have researched, each read from what they are hiring for.

  • Body and Soul Naturally growth strategyA small US maker of natural body care sold through its own Shopify store: aluminum-free deodorant, whipped shea butter, turmeric butter…23 growth ideas
  • Runlayer growth strategyAn AI control plane and enablement platform: a governed MCP gateway and catalog of 18,000+ MCPs, shadow AI discovery, agent identity and…22 growth ideas
  • August Home growth strategyWi-Fi and Bluetooth smart locks that retrofit onto an existing deadbolt, plus keypads, the Connect bridge, Doorsense, doorbell cameras and…24 growth ideas
  • Riboli Family Wines growth strategyA four-generation, family-owned wine and spirits company behind Stella Rosa, Spritz Del Conte, San Antonio Winery, San Simeon, Jada and…22 growth ideas
  • Clara growth strategyAI-powered primary care. Clara imports a member's records from 150,000+ hospitals, labs and pharmacies, chats with them about their…22 growth ideas
  • Mercura growth strategyAI quote and order automation for technical suppliers: Mercura reads emails, bills of quantities, GAEB tenders and purchase orders…23 growth ideas
  • Omni growth strategyAn AI analytics and business intelligence platform built on a semantic model: dashboards, spreadsheets, SQL, point-and-click exploration…21 growth ideas
  • Suzhoudress growth strategyAn online store for made-to-measure occasion dresses: wedding, bridesmaid, mother of the bride, prom, evening, quinceañera and flower girl…23 growth ideas

See every playbook and website review

Independent research. Programmable wrote this playbook from public sources, as they stood on November 10, 2026, and cites them throughout. RunAnywhere did not commission, review or endorse it, and Programmable is not affiliated with RunAnywhere. Company and product names are trademarks of their owners, used only to identify them. The ideas are suggestions and opinion, not professional, legal or financial advice, and not statements about how RunAnywhere runs its business.

Something wrong, or do you represent RunAnywhere? Email [email protected] and we will correct it or take it down.

30/60/90 roadmap

What to focus on in each month, and the ideas that carry it.

  1. Days 1 to 308 ideas

    Make the start visible: $5 credit and prices on the Get started path, a real /pricing page and docs pricing, one engine name, the newsletter archive, the cost chart on the homepage, Wally links in the SDK content, a disclosure on the round-ups and the small speed fixes.

  2. Days 31 to 6012 ideas

    Turn content and integrations into usage: refreshed tutorials, a published SDK licence, self-serve engine evaluation, a public benchmark, model pages, low-balance protection, credit packs, a data-retention page, Wally proof, source tracking, referrals and a launch rhythm.

  3. Days 61 to 904 ideas

    Commit to the big bets: one call that runs on the device or falls back to Wally, dedicated boxes for teams that need their own, an access-grant programme, and Eve as a free public tool.

The research behind the plan

3 research areas, each summarised in a few lines. Open any of them for the full detail.

Customers

What customers say

Customers value low per-token prices for strong open models. The most cited problem: not knowing what a first run costs before signing up.

  • One OpenAI-compatible key and a CLI that plugs into coding agents
  • Agent loops stopping when credit runs out
  • In their words: “open models”

Market

Competitive landscape

4 competitors compared: OpenRouter, Together AI, Fireworks and Cactus.

  • Gap to exploit: One provider that runs the same app on the device and in the cloud
  • OpenRouter: A router, not an engine: speed depends on the provider behind it
  • Together AI: Broad platform rather than a simple agent-first start

Hiring signals

RunAnywhere is a five-person team with one open role, and it is a marketing one: a Developer Relations & Technical Content Intern to write tutorials, integration guides and benchmark write-ups, make demos and build a…

  • Developer Relations & Technical Content Intern: Tutorials, benchmarks, demos and community as the growth engine. Ideas 4, 6, 7, 9, 12, 13, 20.

From the team that wrote this playbook

All products
  • ProgrammableAgencyWorking AI systems in days, not quarters (opens programmable.agency in a new tab)
  • ProgrammableInfoAn agent that finds, researches and asks firstNew accounts get 100,000 free enrichment credits with code BETA100K (opens programmable.info in a new tab)
  • ProgrammableDigitalScreenshots in, an animated product demo out (opens programmable.digital in a new tab)
  • ProgrammableWatchKnow the day a supplier falls out of compliance (opens programmable.watch in a new tab)
  • ProgrammableTalkGet registrants for your events, and know who’s coming (opens programmable.talk in a new tab)
  • ProgrammableMediaAn AI note taker, interpreter and webinar host for your calls (opens programmable.media in a new tab)
  • ProgrammablePageText what you need, your customer gets a page that works (opens programmable.page in a new tab)
  • ProgrammableCommunityForums with labelled AI members, never empty (opens programmable.community in a new tab)
ProgrammableMarketing
ExamplesLeaderboardsProductsTermsPrivacyCookiesDPALegalContact

© 2026 AI-Powered Growth Intelligence