We pick real problems, build focused products, put them in front of the people they are for, and keep the ones that earn their place. Each product has its own audience, roadmap and business model.
Product intelligence · Live
Product discovery from real evidence.
For product managers and founders who run discovery.
Connected to PMEval: Measures the kind of work Praxiom automates.
Praxiom turns customer interviews, support tickets, surveys and call recordings into scored insights, prioritised opportunities and PRDs that link back to the quotes behind them.
Who it’s for
Product managers and founders who run discovery.
Every recommendation traces back to the user quotes that support it.
Drafts PRDs through a draft, review and approve workflow.
Turns approved work into scoped tickets for Linear, GitHub or Jira.
Evidence
Insight
PRD
Ticket
Interview“I gave up when it asked me to invite my team.”
Support ticket“Can I skip adding teammates for now?”
Survey“Setup took longer than I expected.”
Insight · scoredNew users drop off at the team-invite step3 linked quotes
PRD blockMake the invite step skippableevidence: 3
TicketAdd “Skip for now” to invitewhy? → 3 quotes
Every ticket traces back to the evidenceIllustration with example data
Connected research: PMEval. Measures the kind of work Praxiom automates.
agentFast sits above whichever agent framework you already use and adds what production needs: durable execution, human approval before risky actions, memory, guardrails, evals, tracing and live streaming.
Who it’s for
Developers and teams putting agents in front of real customers.
Works with LangGraph, CrewAI, the Claude Agent SDK, the OpenAI Agents SDK, plain Python, or its own multi-agent teams.
A run paused for approval survives a hard kill and resumes from the exact step. Tests check this against a real database on every change.
Source-available and self-hosted: buy it once and run it on your own infrastructure.
Propel is a growth agent for technical founders. It reads your website and repository, works out who the product is for and what is worth saying, then runs marketing and sales as one loop.
Who it’s for
Technical founders who built the product and have no one to sell it.
Availability
Not publicly available yet. Ask about early access and we’ll be in touch when it opens.
Says only what it can prove: every claim traces to your own site, repo, docs or changelog.
Anything that publishes, sends or spends waits for your approval until that channel has earned more freedom.
Search articles, AI-answer visibility, social, launches, email and outbound run in one loop with one memory.
ProposeDraftPublish unless stoppedPublish on its own
Search articles
AI-answer pages
Founder social
Lifecycle email
Outbound
Every claim checked against your site, repo, docs or changelogYour ceiling
Autonomy earned per channel, up to your ceilingIllustration with example channels
Connected research: Propel Gym. Designed to measure Propel’s agent before any real market sees it.
How a product moves through the studio
The same five steps for every product. The last one is a real option.
Find a real problem
Start from a job people already struggle with, not from a model looking for a use.
Build the smallest useful product
Ship something one audience can use for one job, end to end.
Launch to real users
Put it in front of the people it is for, with its own site and its own name.
Learn from usage
Measure what people actually do, and where the product or the AI falls short.
Improve, or stop
Keep investing in what earns its place. Stop what doesn’t, and say so.
What every product has in common
Focused products
Each product does one job for one audience. Breadth comes from the portfolio, not from any single product.
Independent identity
Every product has its own site, roadmap and business model. Einstein Labs is the parent, not a shared login.
Evidence before claims
We show what a product does and who it is for. We don’t publish numbers, logos or testimonials we can’t back.
People stay in control
Where an AI can act, risky actions wait for a person’s approval, and the record of what it did stays visible.
Reliability and cost are features
Products have to survive crashes, bad inputs and real bills, not just demos.
Durable utility
We build for problems that will still exist after the next model release.
The research behind the products
NeoLab works on how to measure agents honestly and keep them reliable in production.