# Bespoke Labs Funding: $40M to Build AI Agent Training Environments > Bespoke Labs raised $40 million across a Series A and seed round in July 2026, backed by Wing VC, 8VC, Jeff Dean and insiders from Anthropic, OpenAI and Meta. Here's the full breakdown. _[Wortins Blog](https://www.wortins.com/blog) · Published Wednesday, July 8, 2026_ **Bespoke Labs has raised $40 million total — a $31.75 million Series A led by Wing VC, layered on top of an earlier $8.25 million seed round led by 8VC** — announced July 6, 2026. The round pulled in an unusually dense bench of AI insiders, including Google DeepMind chief scientist Jeff Dean and angel investors from Anthropic, OpenAI and Meta, to fund a company betting that the bottleneck in AI agents isn't the model — it's what the model practices on. Here's what Bespoke Labs actually builds, who's backing it, and why the "environments" thesis is gaining traction right when agent reliability has become the industry's biggest unsolved problem. ## What is Bespoke Labs? Bespoke Labs is a **Mountain View-based AI research lab that builds training and evaluation environments for autonomous AI agents**. Founded in 2024 by **Mahesh Sathiamoorthy (CEO)** and **Alex Dimakis (Chief Science Officer)**, the company's core insight is that today's agents fail not because the underlying models are weak, but because they've never practiced in anything resembling a real job. Concretely, Bespoke Labs designs simulated environments that look and behave like actual companies: **large codebases, microservices, realistic system logs, support tickets, email threads and Slack-style chat**. Agents are dropped into these environments to complete **long-horizon, economically meaningful tasks** — not toy prompts — and the resulting interaction data feeds back into reinforcement learning pipelines that make the agents measurably more reliable. The company already has a track record in this niche. It built **Terminal-Bench**, a benchmark for evaluating agents on real terminal-based engineering tasks, and released **OpenThoughts**, a reasoning dataset that has been downloaded more than 500,000 times and drawn contributions from engineers at Meta and Amazon. That combination — a widely used benchmark plus a widely used open dataset — is what gave Bespoke Labs credibility with investors before it ever pitched a data-curation-as-a-service business. ## The raise: $40M across seed and Series A The headline numbers: - **Total raised:** $40 million - **Series A:** $31.75 million, led by Wing VC - **Seed round:** $8.25 million, led by 8VC - **Announced:** July 6, 2026 - **Valuation:** not disclosed Unlike many funding stories that lead with one clean headline round, Bespoke Labs' $40 million is a combined figure spanning two rounds — the seed disclosed alongside the new Series A rather than announced separately at the time. That's a common pattern for AI startups that raised quietly and only went public with financials once the bigger round closed, and it's worth flagging for anyone tracking the *pace* of AI funding rather than just the totals. ## Who invested in Bespoke Labs? The **Series A was led by Wing VC**, with participation from **Mayfield**, **The House Fund**, and dbt Labs CEO **Tristan Handy**. The earlier **seed round was led by 8VC**, joined by a notably technical investor group: **Jeff Dean** (Google DeepMind's chief scientist), **Spiros Xanthos** (CEO of Resolve AI) and **Dheeraj Pandey** (CEO of DevRev). Rounding out the cap table are **angel investors from Anthropic, OpenAI and Meta** — meaning people who build frontier models personally chose to bet on a company whose entire product is making agents built on those models actually work. That investor composition is the real signal here. When operators from the three biggest frontier labs write personal checks into an infrastructure layer built *around* their own models, it's a tell that the reliability gap in agents is viewed as structural, not something the next model release quietly fixes. ## What the money is for Bespoke Labs says the capital will go toward three things: 1. **Expanding the research team** — hiring to build out more environment types and RL techniques. 2. **Scaling environment-building infrastructure** — the actual simulated codebases, services and data pipelines agents train against. 3. **Accelerating business momentum** — turning research credibility (Terminal-Bench, OpenThoughts) into paying enterprise relationships. As CEO Mahesh Sathiamoorthy put it, environment infrastructure is "the critical piece needed to optimize and develop agents" — a direct statement of the company's bet that data and environment quality, not raw model scale, is now the limiting factor in agent performance. ## Why it matters Bespoke Labs' raise is a useful marker for where AI investment is heading in the second half of 2026: 1. **The stack is moving up from models to environments.** Foundation model training has largely consolidated around a handful of labs with enough capital to compete. The next layer of value creation — and venture interest — is shifting to the data, evaluation and reinforcement-learning infrastructure that determines whether agents built *on* those models actually work in production. 2. **Insider capital is a credibility signal, not just money.** Angels from Anthropic, OpenAI and Meta, plus Jeff Dean directly, indicates that people closest to frontier model development see agent reliability as an open problem worth funding externally rather than something solved in-house. 3. **Benchmarks and open datasets are becoming go-to-market assets.** Terminal-Bench and OpenThoughts gave Bespoke Labs distribution and credibility before it needed to sell anything — a playbook other infrastructure-layer AI startups are likely to copy. For a company operating one level removed from the foundation models everyone else covers, Bespoke Labs' $40 million says something bigger than its size suggests: reliability, not raw capability, is the next fight in AI, and the smart money is placing multiple bets on who wins it. For more on how this raise stacks up against the rest of the market, see Wortins' running list of the [biggest AI funding rounds of 2026](/blog/biggest-ai-funding-rounds-2026). --- *Following AI funding? Wortins tracks the biggest raises, valuations, and acquisitions daily in the [AI Funding Tracker](/funding).* ## Frequently asked questions ### How much did Bespoke Labs raise? Bespoke Labs raised $40 million total: a $31.75 million Series A led by Wing VC and an earlier $8.25 million seed round led by 8VC. ### Who invested in Bespoke Labs? Wing VC led the Series A with Mayfield, The House Fund and dbt Labs CEO Tristan Handy participating. 8VC led the seed round alongside Jeff Dean, Spiros Xanthos and Dheeraj Pandey, plus angel investors from Anthropic, OpenAI and Meta. ### What does Bespoke Labs do? Bespoke Labs builds realistic training and evaluation environments — simulated codebases, microservices, tickets, logs and chat systems — that let AI agents practice long-horizon, economically meaningful work before they're deployed in production. ### Who founded Bespoke Labs? Bespoke Labs was founded in 2024 by Mahesh Sathiamoorthy (CEO) and Alex Dimakis (Chief Science Officer), and is based in Mountain View, California. --- _Curated and written by [Wortins](https://www.wortins.com) — The daily AI briefing. Every story links to its original source; the "Wortins read" on each is our own original analysis. [About Wortins & our editorial approach](https://www.wortins.com/about)._