Remanence, a startup founded by alumni of France’s H Company, has raised a $30 million seed round to turn AI agents’ work into training data for companies’ own models. FirstMark said the round was co-led by FirstMark and EQT Ventures, “with participation from Zero Prime, Emblem, Drysdale, Global Founders Capital, and Yellow”.
Our number: the round is 4 times the $7.5 million seed that Unsiloed AI announced this week, by our arithmetic from the two announcements. FirstMark names seven investment firms in the round, plus individual angels, by our count.
Business Pill · Vertical Integration
An evergreen Business Engineer explainer on vertical integration. It teaches the general idea only and does not discuss Remanence.
The key insight: Remanence raised a $30 million seed co-led by FirstMark and EQT Ventures to turn companies’ AI-agent traces into reinforcement-learning environments and post-train open-weight models that the customer owns, 4 times Unsiloed’s seed by our arithmetic.
What Remanence Does
FirstMark says “Remanence plugs into a company’s existing AI models and turns their execution traces into tailored training environments, then post-trains open-weight models through reinforcement learning”.
The company’s own site describes four steps, starting with trace analysis: “Connect your observability stack. Find where your agent fails.” It describes the training step as “Any open-weight model, trained against your live baseline.”
FirstMark says ownership is the pitch: “Each customer ends up with a specialized model of its own, the company keeps the model weights, and its data can stay inside its own infrastructure”.

The Team and the Money
FirstMark says co-founders Pierre-Louis Cedoz, Hamza Benchekroun and Axel Moyal previously worked together at H Company, “France’s computer-use AI lab”, where they led the post-training of the open-weight Holo models. It says the company is “working with customers in legal, customer support, and enterprise software across Europe and the US”.
FirstMark says “The new funding will support research and compute infrastructure, including a dedicated GPU cluster”. It says the team is hiring researchers and engineers in Paris, London and San Francisco.
The bet sits on open-weight models, the same layer where Chinese labs now lead usage rankings; this week Washington weighed making US companies disclose their use of Chinese models.
The Structural Read
As we read FirstMark’s announcement, the product is the data agents already throw away: it says almost all execution data is discarded today.
As we read the two pitches together, Remanence sells model ownership, with customers keeping the weights, where most enterprise AI is rented by the token.
FirstMark, on why it co-led the round
“Rather than renting more general-purpose intelligence every quarter, companies can build models that get better the more they’re used.”
Three Implications
FOR COMPANIES RUNNING AGENTS FirstMark says each customer keeps the weights of its own specialised model.
FOR OPEN-WEIGHT MODEL MAKERS Remanence says it trains any open-weight model against the customer’s live baseline.
FOR INVESTORS FirstMark names five participating firms alongside the two co-leads, plus individual angels.
The Business Engineer Lens
This story maps onto the Business Engineer framework The Agentic Harness War.
The analysis puts it this way: “The surface, not the model, governs autonomy.”
As we read FirstMark’s announcement, Remanence bets on the other side of that line: that training on the surface’s own traces is what makes the model worth owning. The analysis predates this round and does not discuss Remanence.
What Is Not Established
FirstMark’s announcement does not give a valuation, revenue, customer names or the split between the two lead investors. Remanence does not say which open-weight models it trains most.
We did not contact Remanence, FirstMark or EQT Ventures.
The Bottom Line
Remanence raised a $30 million seed, co-led by FirstMark and EQT Ventures, to post-train open-weight models on companies’ own agent traces. The round is 4 times Unsiloed’s seed announced this week, by our arithmetic.
97,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.
A note on sourcing. The round, investors, product, team and use of funds are from FirstMark’s announcement of 17 November 2026 and Remanence’s website, both read in full; Unsiloed’s $7.5 million seed is from our 18 November piece. The 4x comparison and the investor count are our arithmetic. We did not contact the companies.
Sources: FirstMark, ‘Why We Co-Led Remanence’s $30M Seed’, 17 Nov 2026 · Remanence website








