Cinder's new AI Growth Marketer is asked for "original research or benchmarks that set Cinder apart", and Cinder already has them: the October 21 test of six open-weight models (four gave harmful help in over 40% of responses) and the August test of an abliterated Alibaba model (23.1% violation rate across 2,670 attacks). But the results sit in the same list as hiring posts, end only in a newsletter box, and in the Alibaba post the two places where tables belong show a raw "Unknown block type tableBlock" developer message instead.
First action
Render the two tables in the Alibaba post today and check the other posts for the same message.
How to measure it
Red-team scoping calls from research
- Now
- –
- Target
- 10 a month
- By
- 90 days
Steps
Fix the template
Add the table component and re-publish the Alibaba post.
Web developer 1 day
Check every post
Search all 59 posts for unrendered blocks.
Web developer 1 day
Build /research
One card per test: model, date, method, headline rate.
Senior Product Designer 1 week
Make the download
One-page PDF and category list per test, behind a short form.
AI Growth Marketer 1 week
Add the call to action
"Test your model" books a scoping call with sales.
Head of Sales 2 days
Brief press and labs
Send the series page to safety reporters and model teams.
AI Growth Marketer 1 week
Set a cadence
One new test each month, announced in the newsletter.
Red team lead monthly
Measure
Downloads, calls booked and pipeline by test.
AI Growth Marketer monthly
Why it works
Fix the table block in the blog template and re-check every post. Then give the research its own home: a /research page that lists each test with the model, date, method, headline rate and a chart; a one-page PDF and the prompt categories (not the prompts) behind a short form; and a "test your model" call to action that books a red-team scoping call. Publish each new test as part of the series, so press, AI labs and safety teams start to expect the next one.
Generative AI buyers trust numbers from tests they could run themselves. A named, recurring series is what makes a one-off blog post into a demand asset, and the form turns readers into the pipeline the role is measured on.
Ready to use
ChecklistResearch series launch list
[ ] Table block renders in the blog template [ ] No raw developer messages in any post [ ] /research page with one card per test [ ] Method, date and model on every card [ ] One-page PDF per test behind a short form [ ] "Test your model" call to action to a scoping call [ ] Newsletter issue for each new test [ ] Pipeline tracked by test
Risks and how to handle them
Model makers object to results
Publish the method, share findings with the model team before release, and stick to measured rates.
Forms cut readership
Keep the article open; gate only the PDF and category list.
Evidence
- Your websiteRed-team post "What Cinder Learned Red Teaming Alibaba's Latest Open-Weight Model" (August 24, 2026): across 2,670 attacks the original model generated NCII once and the abliterated version violated 23.1% of the time. Two places in the post where tables belong show the text "[@portabletext/react] Unknown block type "tableBlock", specify a component for it in the `components.types` prop" instead (checked twice on October 22, 2026). The post ends with a newsletter sign-up; there is no download of the data or method.Source
- Your website"How Leading Open-Weight Models Respond to Requests Involving Violent Harm" (October 21, 2026): six open-weight models tested against 209 adversarial requests built from 70 goals, 18 attack techniques and 15 harm domains; "Four of the six gave harmful, actionable help in more than 40% of their responses, and the worst reached 56%. Even the safest gave unsafe responses about 15% of the time."Source
- HiringAI Growth Marketer duties: lift the demand-gen performance of case studies ("One of our current case studies is already one of our best-performing assets... good enough that customers ask to be featured"); turn the blog, newsletter and founder ghostwriting into "an explosive growth channel" with "original research or benchmarks that set Cinder apart"; decide which events to attend, sponsor or skip; "Spot and act on relevant news and industry moments quickly"; own "attribution, lead scoring, and inbound reporting"; and partner with sales "so marketing-sourced leads actually convert".Source
- HiringAI Growth Marketer requirements: demand generation "for an AI product specifically" is "the requirement we care about most"; built from zero; experience in a category "where the downside of failure is existential (cybersecurity, identity and authentication, fintech risk, regtech...)". Nice to have: "You've built a research report, index, or open benchmark that became a demand asset", "ABM or intent-data tools", "Field marketing at conference scale", and marketing into "social platforms, marketplaces, dating apps, gaming, fintech apps, or B2C-facing AI products".Source
- Your websiteResources hub: 65 items (Insights 37, Product 10, Engineering 7, News 3, Case studies 4); a newsletter sign-up "sent only when we have something worth your attention". The red-team research sits in the same list as hiring and culture posts, with no research or benchmark section.Source