Rerun growth playbook

Publish one /community page: upcoming events (pulled from the Luma calendar), past events with the talk slides, notebook and recording, a "bring Rerun to your city or lab" form that asks city, team and what they are building, and links to Discord, GitHub and the Learn course. Point every event page, the footer and the Discord welcome message at it. Every form answer lands in the commercial stack the Founder's Office is building, tagged by city and company.

A recurring program needs a home that keeps value after the event, and the "your city" form turns scattered interest into a ranked list of where to go next and which teams to talk to about Hub.

[ ] Upcoming events from Luma
[ ] Past events: slides, notebook, recording
[ ] "Bring Rerun to your city or lab" form
[ ] Form entries routed to the CRM with city and company
[ ] Linked from footer, event pages, Discord welcome
[ ] Release-notes email mentions the next event
[ ] Monthly review: next three cities

• An empty calendar looks worse than none

  Launch with past events and recordings, and at least two dated upcoming events.

• Sign-ups without follow-up

  Agree who answers each Hub-interested entry within two working days.

• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source
• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source
• Your websiteROSCon page: "Thanks for coming to our talk!", an after-party meetup, and a tidy set of next steps (open an MCAP bag, record ROS 2 bags as .rrd, query with dataframes, stream to PyTorch, "Export to LeRobot"). The Munich and Zurich meetup pages follow the same pattern with "Book a meeting", Discord and GitHub links; registration happens on Luma.Source

Rewrite the blog index description around the data layer for physical AI, Hub and the Spotlight series. Check the descriptions of the older top posts (Why Rust, release posts) and add a one-line note on old release posts pointing to the current changelog.

Search previews and AI summaries repeat meta descriptions, so the old line keeps teaching people that Rerun is only a viewer.

• Old posts lose their search terms

  Keep "visualization" in the description; add the data layer, do not remove the viewer.

• Your websiteBlog ("News"): its meta description still reads "Rerun is a visualization toolbox for computer vision", while every other page says "The Data Layer for Physical AI". Release-highlight posts stop at 0.27, while rerun-sdk 0.40.0 shipped on PyPI on October 21, 2026 (release notes now live in the docs changelog). Recent posts are a strong "Spotlight" series on ARKitScenes depth, labels and 3D reconstruction.Source
• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source
• HiringSeven open roles on Rerun's Ashby board: the three above plus Rust and infrastructure engineers and a Talent Partner (Technical), posted October 14, 2026. The community post compares Rerun's strategy with "the open source to commercial platform playbook of Confluent, MongoDB, and Elastic, except Rerun is building the category as well as the company", and names teams at "Meta, Google, Hugging Face, and Unitree" as users.Source

Replace the free text with five fields before the calendar: team size, data volume (GB, TB, PB), formats (MCAP, ROS bags, LeRobot, custom), where the data lives (S3, GCS, on-prem) and timeline. Route PB-scale or near-term teams straight to a founder or forward-deployed engineer, academic labs to the research path, and small teams to the SDK course with an invitation to come back.

Structured answers let a tiny team spend founder time on the deployments that fit, and make the first call a technical one instead of discovery.

• Fewer bookings

  Keep the form under a minute; make every field a click, not typing.

• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source
• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source
• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source
• Your websiteContact page: Hub, pricing and design-partnership enquiries go to a contact form ("Include the scale of your recordings, the data formats you use, and the outcome you want"), technical questions to Discord, bugs to GitHub. llms.txt describes a local "rerun viewer-mcp" command that lets an agent "inspect and control a running Rerun Viewer"; it is documented in the CLI manual rather than on its own docs page.Source

Within 48 hours of each event, send attendees one email from the speaker: the recording, the notebook from the talk, the matching Learn chapter, the Discord channel, and one line inviting teams with large recordings to a Hub walk-through. Log who opens the notebook so the Founder's Office sees which teams are engaged.

Interest fades within days; a useful follow-up from the person who spoke converts goodwill into installs and Hub conversations.

• Feels like a sales email

  Lead with the notebook; keep the Hub line to one sentence.

• Your websiteROSCon page: "Thanks for coming to our talk!", an after-party meetup, and a tidy set of next steps (open an MCAP bag, record ROS 2 bags as .rrd, query with dataframes, stream to PyTorch, "Export to LeRobot"). The Munich and Zurich meetup pages follow the same pattern with "Book a meeting", Discord and GitHub links; registration happens on Luma.Source
• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source
• Your websiteMedia kit: the SDK "is used by teams at Hugging Face, DeepMind, NVIDIA, and others working at the frontier of Physical AI"; co-founders Nikolaus West (CEO), Emil Ernerfeldt (CTO, creator of egui) and Moritz Schiebold (COO, formerly CEO of Volumental). The Learn course takes "a bimanual robot-arm dataset" from raw MCAP logs to LeRobot training data in six articles.Source

Add a short, factual panel to the 15 docs pages about querying many recordings, serving datasets and training dataloaders: what changes at scale, a link to the DROID Hub video, and the qualifier form. Keep the open source docs complete; the panel only points to the next step.

Developers decide to scale while reading docs; meeting them there turns the open source funnel into Hub pipeline without changing the free product.

• Open source users dislike sales in docs

  Write the panel as a technical note, never a pop-up.

• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source
• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source
• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source

Publish one post a month with the three biggest changes, a GIF each and an upgrade link, and send it to the release-notes list. Link the changelog for detail.

Visible momentum helps open source adoption, recruiting and Hub buyers who check whether a vendor is active.

• Repeats the changelog

  Pick only user-visible wins and show them.

• Your websiteBlog ("News"): its meta description still reads "Rerun is a visualization toolbox for computer vision", while every other page says "The Data Layer for Physical AI". Release-highlight posts stop at 0.27, while rerun-sdk 0.40.0 shipped on PyPI on October 21, 2026 (release notes now live in the docs changelog). Recent posts are a strong "Spotlight" series on ARKitScenes depth, labels and 3D reconstruction.Source
• GitHubrerun-io/rerun has 11,627 stars and was pushed to on October 22, 2026; the organization has 148 public repositories and 18 active in the last 90 days. The Discord is linked from every page. The product page says "10K+ GitHub stars".Source

Write four stories with the open source teams (how they use Rerun, what it replaced, what it saved), then two Hub stories with early customers who agree, even anonymised by industry. Publish a /customers page and use the stories in Hub meetings and events.

Physical AI teams buy what peers use; named stories turn the logos into proof and give the community manager talk material.

• Hub customers will not be named

  Publish by industry with measured results.

• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source
• Competitor researchRivals' own homepages: Foxglove, "Agentic data platform for Physical AI", "Free to start", with Customers, Customer stories and seven industry pages (automotive, defense, manufacturing, logistics, agriculture, marine, construction); Roboto, the "Analytics Engine for Robotics & Physical AI", with Customers and Industries menus, AI agents for robotics data and a Gold sponsorship of ROSCon 2026.See the research Source
• HiringForward Deployed Engineer (remote US West Coast): "Rerun Hub, our data lakehouse, is in private preview and already managing petabytes of data for early customers"; the engineer works "side by side with our customers over multi-month engagements", runs "technical discovery", and will "Feed recurring issues, patterns, and product gaps back" to product, with about 30% travel.Source
• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source

Build four pages from existing examples, each with the data types, a browser-viewable recording, the training path (LeRobot export or dataloader) and the Hub angle. Link them from the homepage and the docs panel.

Use-case pages rank for problem searches and let a buyer see their own data shape before a call.

• Thin pages

  Each page needs a real recording and code.

• Competitor researchRivals' own homepages: Foxglove, "Agentic data platform for Physical AI", "Free to start", with Customers, Customer stories and seven industry pages (automotive, defense, manufacturing, logistics, agriculture, marine, construction); Roboto, the "Analytics Engine for Robotics & Physical AI", with Customers and Industries menus, AI agents for robotics data and a Gold sponsorship of ROSCon 2026.See the research Source
• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source
• Your websiteMedia kit: the SDK "is used by teams at Hugging Face, DeepMind, NVIDIA, and others working at the frontier of Physical AI"; co-founders Nikolaus West (CEO), Emil Ernerfeldt (CTO, creator of egui) and Moritz Schiebold (COO, formerly CEO of Volumental). The Learn course takes "a bimanual robot-arm dataset" from raw MCAP logs to LeRobot training data in six articles.Source

Write three fair guides: from RViz, from rosbag scripts, and from Foxglove, each mapping concepts, showing the same bag in Rerun, and listing what Rerun does not do. Add a short comparison table.

Switching searches are high intent, and the ROS material already exists to build on.

• Rival claims go stale

  Date each guide and cite the rival's own docs.

• Your websiteROSCon page: "Thanks for coming to our talk!", an after-party meetup, and a tidy set of next steps (open an MCAP bag, record ROS 2 bags as .rrd, query with dataframes, stream to PyTorch, "Export to LeRobot"). The Munich and Zurich meetup pages follow the same pattern with "Book a meeting", Discord and GitHub links; registration happens on Luma.Source
• Competitor researchRivals' own homepages: Foxglove, "Agentic data platform for Physical AI", "Free to start", with Customers, Customer stories and seven industry pages (automotive, defense, manufacturing, logistics, agriculture, marine, construction); Roboto, the "Analytics Engine for Robotics & Physical AI", with Customers and Industries menus, AI agents for robotics data and a Gold sponsorship of ROSCon 2026.See the research Source
• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source

Publish a research page: free or discounted Hub for labs, a citation snippet, and a list of papers and datasets that use Rerun (searchable on arXiv and GitHub). Invite authors to a monthly research talk.

Citations compound: every paper that uses Rerun recruits the next lab, and the list is proof no competitor can copy quickly.

• Academic usage costs Hub capacity

  Cap storage per lab.

• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source
• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source
• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source

Build a calculator: GPUs, hours idle waiting for data per training run, engineers maintaining export jobs, and data volume. It shows the yearly cost of the current setup and passes the answers into the qualifier form.

A buyer who has put a number on the problem arrives at the first meeting ready to talk scope, and the answers give sales the business case.

• Inflated numbers

  Show the method and use ranges.

• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source
• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source
• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source

Each week, combine company-level signals that people make public (GitHub organizations opening issues or PRs, Discord members who state their company, event sign-ups and docs-panel clicks) into one ranked account list. Founders reach out to the top five with something useful, never a cold pitch.

In open source companies, the best Hub prospects are teams already struggling with scale on the free tool.

• Community feels watched

  Use only public company-level signals and lead with help.

• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source
• GitHubrerun-io/rerun has 11,627 stars and was pushed to on October 22, 2026; the organization has 148 public repositories and 18 active in the last 90 days. The Discord is linked from every page. The product page says "10K+ GitHub stars".Source
• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source

Run a monthly evening in each city with one format: two short talks on real data problems (one from a user), a live debugging session on an attendee's recording, and food. Record every talk to YouTube and post it on /community.

A fixed format and date builds a habit and a content library, and the live debugging session shows the product at its best.

• Speaker supply

  Ask every Hub customer and open source team for one talk a year.

• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• HiringSeven open roles on Rerun's Ashby board: the three above plus Rust and infrastructure engineers and a Talent Partner (Technical), posted October 14, 2026. The community post compares Rerun's strategy with "the open source to commercial platform playbook of Confluent, MongoDB, and Elastic, except Rerun is building the category as well as the company", and names teams at "Meta, Google, Hugging Face, and Unitree" as users.Source
• Your websiteSitemap (691 URLs, read October 22, 2026): 528 docs pages, 100 examples, 35 blog posts and a 7-part Learn course; three event pages (/roscon for the ROSCon Global 2026 talk in Toronto, and meetup pages for Munich and Zurich). There is no events or community index, no customer or case-study page, and no use-case or industry pages (/events, /community, /customers and /case-studies return 404).Source

After each deployment, write an anonymised reference architecture: data sources, storage, ingest, queries, training path, and lessons. Publish them in docs and share them in Hub meetings.

Architectures reduce buyer risk and shorten technical discovery for the next customer.

• Customer sensitivity

  Remove names and scale details on request.

• HiringForward Deployed Engineer (remote US West Coast): "Rerun Hub, our data lakehouse, is in private preview and already managing petabytes of data for early customers"; the engineer works "side by side with our customers over multi-month engagements", runs "technical discovery", and will "Feed recurring issues, patterns, and product gaps back" to product, with about 30% travel.Source
• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source
• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source

Create a contributors page, monthly shout-outs in the release post, early access to Hub features, and an annual contributor day with the core team.

Recognised contributors stay longer, write integrations and become local meetup hosts.

• Favours a few people

  Publish clear criteria.

• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• GitHubrerun-io/rerun has 11,627 stars and was pushed to on October 22, 2026; the organization has 148 public repositories and 18 active in the last 90 days. The Discord is linked from every page. The product page says "10K+ GitHub stars".Source
• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source

Host a read-only Hub with DROID and one or two other public datasets. Anyone with a GitHub login can run SQL and dataframe queries and stream a slice to a notebook, with limits. Offer a meeting when they hit the limit.

Trying beats watching for developers, and sandbox usage tells sales who is serious.

• Cost of free compute

  Read-only, rate limited, one region.

• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source
• Competitor researchRivals' own homepages: Foxglove, "Agentic data platform for Physical AI", "Free to start", with Customers, Customer stories and seven industry pages (automotive, defense, manufacturing, logistics, agriculture, marine, construction); Roboto, the "Analytics Engine for Robotics & Physical AI", with Customers and Industries menus, AI agents for robotics data and a Gold sponsorship of ROSCon 2026.See the research Source
• Your websitePricing page: the SDK is "Free forever" (Apache-2.0 / MIT); Hub is "Contact us", "priced per-deployment based on team size and data scale", single-tenant in the customer's region over their own S3-compatible buckets, with support "from a shared Slack channel to a dedicated forward-deployed engineer". The page says Hub "replaces the export pipelines, custom dataloaders, and idle-GPU time you pay for today" and that Rerun offers academic and research discounts on request.Source

Publish a /datasets gallery: popular public datasets (ARKitScenes, DROID, LeRobot datasets) as browser-viewable recordings with a one-line load snippet, a page per dataset and a "train from this" example.

Dataset-name searches bring exactly the right engineers, and each page shows the viewer and query tools on data they already care about.

• Licence limits

  Link to the source where redistribution is not allowed.

• Your websiteBlog ("News"): its meta description still reads "Rerun is a visualization toolbox for computer vision", while every other page says "The Data Layer for Physical AI". Release-highlight posts stop at 0.27, while rerun-sdk 0.40.0 shipped on PyPI on October 21, 2026 (release notes now live in the docs changelog). Recent posts are a strong "Spotlight" series on ARKitScenes depth, labels and 3D reconstruction.Source
• Your websiteHub page: "Petabyte-scale", "Direct streaming", "Built for physical data"; "The quickest way to get Hub is to watch it": a five-minute video of a notebook exploring the DROID dataset, with the same query scaling "from a handful of clips to 70,000 across terabytes". The only call to action is "Book a meeting".Source
• Your websiteMedia kit: the SDK "is used by teams at Hugging Face, DeepMind, NVIDIA, and others working at the frontier of Physical AI"; co-founders Nikolaus West (CEO), Emil Ernerfeldt (CTO, creator of egui) and Moritz Schiebold (COO, formerly CEO of Volumental). The Learn course takes "a bimanual robot-arm dataset" from raw MCAP logs to LeRobot training data in six articles.Source

Propose a joint program: a shared tutorial series, a joint hackathon, a co-hosted dataset challenge, and a Hub offer for LeRobot dataset publishers.

Partnering with the most visible open robot learning project borrows its reach and makes Rerun the default data path.

• Partner priorities change

  Start with one tutorial and grow.

• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source
• Your websiteROSCon page: "Thanks for coming to our talk!", an after-party meetup, and a tidy set of next steps (open an MCAP bag, record ROS 2 bags as .rrd, query with dataframes, stream to PyTorch, "Export to LeRobot"). The Munich and Zurich meetup pages follow the same pattern with "Book a meeting", Discord and GitHub links; registration happens on Luma.Source
• Your websiteMedia kit: the SDK "is used by teams at Hugging Face, DeepMind, NVIDIA, and others working at the frontier of Physical AI"; co-founders Nikolaus West (CEO), Emil Ernerfeldt (CTO, creator of egui) and Moritz Schiebold (COO, formerly CEO of Volumental). The Learn course takes "a bimanual robot-arm dataset" from raw MCAP logs to LeRobot training data in six articles.Source

Survey the community (Discord, events, release list) on data volume, formats, storage, tools and time lost. Publish the results with charts, and present them at the big conferences.

Original data earns press and citations and gives every event and sales conversation a new opening.

• Biased sample

  Partner with a university for distribution.

• Your websiteBlog ("News"): its meta description still reads "Rerun is a visualization toolbox for computer vision", while every other page says "The Data Layer for Physical AI". Release-highlight posts stop at 0.27, while rerun-sdk 0.40.0 shipped on PyPI on October 21, 2026 (release notes now live in the docs changelog). Recent posts are a strong "Spotlight" series on ARKitScenes depth, labels and 3D reconstruction.Source
• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• Company blogRerun raised $17 million in seed funding led by Point Nine, with Costanoa, Sunflower Capital, Seedcamp and angels including Guillermo Rauch, Eric Jang, Oliver Cameron, Wes McKinney and Nicolas Dessaigne, "to build out the data stack for Physical AI".Source

Write a dedicated guide and short video: connect an AI coding agent to the Viewer, let it inspect a failing run, and fix the code. Add it to the homepage and docs navigation and post it where agent users gather.

Coding agents are how many engineers now write robotics code; being the agent's eyes on physical data is a new reason to install Rerun.

• Fast-changing agent tools

  Keep the guide tool-agnostic.

• Your websiteContact page: Hub, pricing and design-partnership enquiries go to a contact form ("Include the scale of your recordings, the data formats you use, and the outcome you want"), technical questions to Discord, bugs to GitHub. llms.txt describes a local "rerun viewer-mcp" command that lets an agent "inspect and control a running Rerun Viewer"; it is documented in the CLI manual rather than on its own docs page.Source
• Your websiteHomepage: "The Data Layer for Physical AI", "pip install rerun-sdk", a "Try in browser" link and a "Book a meeting" button for Hub; open source users shown are LeRobot (Hugging Face), Brush (DeepMind), PyCuVSLAM (NVIDIA) and Project Aria (Meta Reality Labs Research), followed by a wall of posts from X, including one saying agentic coding needs "a tight feedback loop with clear error signals".Source

Publish it as an interactive map of labs, meetups and companies by city, open to submissions. Each submission is a contact, and the map shows where to hold the next event.

A useful public resource earns links and gives every lab a reason to talk to Rerun.

• Stale entries

  Review monthly.

• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• HiringStrategic Operations, Founder's Office (Stockholm), "one of our first commercial hires": "Demand is strong" and the role will "set up and automate our go-to-market stack such as lead capture and triage", "Turn community signals into opportunities: dig into activity across our community and developer channels to work out who's engaged, where they work, and when to reach out", "build an annual events calendar", help the founders with content, and "Assess inbound".Source

Add a "we're hiring" slide and page link at every event, tag contributors and frequent Discord helpers for the Talent Partner, and invite candidates to Data Nights.

Candidates who already use the product ramp faster and stay longer.

• Community feels recruited

  Keep it light: one slide and a link.

• HiringSeven open roles on Rerun's Ashby board: the three above plus Rust and infrastructure engineers and a Talent Partner (Technical), posted October 14, 2026. The community post compares Rerun's strategy with "the open source to commercial platform playbook of Confluent, MongoDB, and Elastic, except Rerun is building the category as well as the company", and names teams at "Meta, Google, Hugging Face, and Unitree" as users.Source
• HiringDeveloper Community Manager (hybrid San Francisco, posted August 27, 2026): the physical AI community is "scattered across a handful of hubs and a lot of Discord servers and meetups" and "doesn't have a heartbeat yet, and that's this role". The job is to "Map the physical AI landscape", "Design and run developer events", "Build recurring programs that give the community real content and social touch points, not one-off events that go nowhere" and represent Rerun "with key open source contributors"; the hire will later partner with "Rerun's VP of Marketing in SF once that role is filled".Source
• GitHubrerun-io/rerun has 11,627 stars and was pushed to on October 22, 2026; the organization has 148 public repositories and 18 active in the last 90 days. The Discord is linked from every page. The product page says "10K+ GitHub stars".Source

Defer the X wall and GitHub widgets below the fold, split JavaScript by route, compress images, and set cache lifetimes. Recheck lab and field numbers.

A faster first view helps event and mobile traffic without hurting anything.

• Breaks embeds

  Test the social wall after the change.

• Google PageSpeedHomepage on mobile: performance 62/100; in the lab the main content (the "Open source SDK" heading) takes 8.2s to appear and the page downloads 2.2 MB with 264 KB of unused JavaScript. Real visitors pass Core Web Vitals (field LCP 1.2s, INP 129ms, CLS 0), and field LCP improved 23% on the prior period.Source