The archive · AI & Models · Product decision · 2024–2026
Reflection AI bets coding agents lead to superintelligence: $130M launch, $2B at $8B
Ex-DeepMind researchers launched Reflection AI betting autonomous coding leads to superintelligence; seven months later Nvidia led $2B at an $8B valuation
Reflection AI
What the business is
An AI research startup building autonomous coding agents (first product Asimov) and open-weight frontier models, aiming at AI that performs most computer-based work.
Starting capital：$130M at launch (2025-03); +$2B (2025-10)
How it started
Misha Laskin and Ioannis Antonoglou, former Google DeepMind researchers who worked on Gemini's training workflow and post-training systems, founded Reflection AI and emerged from stealth in March 2025 with $130 million: a $25 million seed led by Sequoia and CRV plus a $105 million Series A led by CRV and Lightspeed, valuing the company at about $555 million. Reported backers included Nvidia's venture arm, Reid Hoffman and Scale AI's Alexandr Wang. The company defined superintelligence as AI that performs most computer-based work, and chose autonomous coding as the first step.
What happened
At launch Reflection already had paying customers in financial services and technology with large coding teams. Its first autonomous coding agent, Asimov, shipped in July 2025, designed to ingest a business's code and data to understand why systems do what they do. On October 9, 2025 the company announced a $2 billion round led by Nvidia at an $8 billion valuation — more than 14 times its March valuation — with Lightspeed, Sequoia Capital Operations, DST Global, Eric Schmidt and 1789 Capital participating. The capital funds open-source frontier models, global infrastructure and safety research.
How it ended up
No exit — still independent and scaling: after the October 2025 round the company is valued at $8 billion, pushing open-source models, Asimov and infrastructure spending.
Background
Reflection AI is an AI startup founded by Misha Laskin and Ioannis Antonoglou, former Google DeepMind researchers who worked on Gemini's training workflow and post-training systems. It emerged from stealth in March 2025 with $130 million — a $25 million seed led by Sequoia and CRV plus a $105 million Series A led by CRV and Lightspeed — valuing the company at about $555 million, with Nvidia's venture arm, Reid Hoffman and Scale AI's Alexandr Wang among reported backers.
The company's thesis is that autonomous coding agents are the first step to superintelligence: systems that read, write, test and deploy code, handling infrastructure on a developer's behalf. At launch it already had paying customers in financial services and technology. Its first agent, Asimov, shipped in July 2025; the company said the breakthroughs needed for fully autonomous coding — advanced reasoning and iterative self-improvement — extend naturally to broader categories of computer work.
On October 9, 2025, Reflection announced a $2 billion round led by Nvidia at an $8 billion valuation, with Lightspeed, Sequoia Capital Operations, DST Global, Eric Schmidt and 1789 Capital participating. The company frames its work as open and auditable — frontier intelligence should not be locked behind closed labs — and is spending the capital on open-source frontier models, global infrastructure and safety research.
What has to be true
- Two researchers with hands-on Gemini training and post-training experience made the autonomous-coding thesis credible enough to raise $130M before shipping
- The company defined superintelligence operationally — AI that performs most computer-based work — and made coding the 'root node' to bootstrap everything else
- Nvidia leading the $2B round at $8B shows chipmakers betting on open, agentic workloads as a new compute market
- Openness is a deliberate contrast to closed labs, giving Reflection a distribution and trust position even while it raises frontier-scale capital
What can be applied
Backing researchers who built a frontier model's training pipeline can outrun product: $130M at launch, then $2B at $8B on the autonomous-coding thesis. Credibility was priced before proof.
Aftermath
As of September 1, 2026, Reflection AI remains private and scaling: its total disclosed funding is over $2.1 billion ($130M at launch plus the $2B Nvidia-led round), with the October 2025 valuation of $8 billion unchanged in public reporting. It continues to develop its Asimov coding agent and open-source frontier models, and to invest in infrastructure and safety evaluation. No IPO, acquisition or new round has been announced, and the open-vs-closed frontier-lab competition it bet on is still unresolved.
Sources
- Superintelligence startup Reflection AI launches with $130M in funding
- Reflection AI Raises $130 Million to Build Autonomous Coding Systems
- Reflection AI lands $2B at $8B valuation to expand frontier AI infrastructure and safety research
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card