Jason Warner
Last reviewed
Jun 8, 2026
Sources
17 citations
Review status
Source-backed
Revision
v1 · 1,896 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 8, 2026
Sources
17 citations
Review status
Source-backed
Revision
v1 · 1,896 words
Add missing citations, update stale details, or suggest a clearer explanation.
Jason Warner is an American software engineer and technology executive who is the co-founder and chief executive officer of Poolside, a company building foundation models for software development. Before co-founding Poolside in 2023 with Eiso Kant, Warner spent roughly two decades in engineering leadership, most prominently as the chief technology officer of GitHub from 2017 to 2021. That period spanned Microsoft's 7.5 billion dollar acquisition of GitHub and the early incubation of GitHub Copilot, the AI pair programmer that helped define the modern market for AI code generation. Warner has also led engineering at Heroku and Canonical and worked as a managing director at the venture capital firm Redpoint Ventures.[1][2][3]
Warner has described himself as a late starter in programming. He has said that he expected to study mechanical or aerospace engineering, with a minor in physics or mathematics, and did not begin writing code until he reached university. He earned a bachelor's degree in computer science from Pennsylvania State University and later completed a master of science degree at Rensselaer Polytechnic Institute.[3][4]
His path into software came through the early consumer internet. He has recounted noticing that his dormitory had network access, experimenting with simple web and back-end servers, and concluding that the web would underpin most future computing. Around 1996 and 1997 he encountered Netscape and the emerging open source community, including the Linux operating system, which drew him toward systems and infrastructure work.[3][4] He later titled a widely shared talk and interview around the idea of "how a very average programmer became GitHub's CTO," using his own career as an argument that engineering leadership depends more on judgment and people than on raw coding talent.[3]
Warner built his early reputation running large engineering organizations rather than as an individual contributor. He spent close to four years at Canonical, the company behind the Ubuntu Linux distribution, where he oversaw product engineering for Ubuntu Desktop and the Ubuntu Phone effort.[2][5]
He then joined Heroku, the cloud platform-as-a-service owned by Salesforce, where he served for more than three years as vice president of engineering, leading the team responsible for the platform that hosted and scaled web applications for many startups.[2][5] These roles gave Warner deep exposure to developer tooling, cloud infrastructure, and the day-to-day workflows of professional software teams, themes that would later shape his work at GitHub and Poolside.
Warner joined GitHub in 2017 and served as its chief technology officer until 2021. During his tenure he led the engineering organization through a period of rapid product expansion, bringing to market features including GitHub Actions for continuous integration and automation, Packages for software distribution, Advanced Security, Connect, and Codespaces, the company's cloud development environment.[1][2][6]
His time as CTO coincided with one of the most significant events in the company's history. In June 2018 Microsoft agreed to acquire GitHub for 7.5 billion dollars in stock, a deal that closed later that year. Warner remained chief technology officer through the transition and helped integrate the company while preserving its developer-focused culture.[6][7] He has spoken publicly about leading engineering teams through that acquisition and other moments of organizational stress.[1]
Warner is also closely associated with the early development of GitHub Copilot. Under his engineering leadership, GitHub partnered with OpenAI to incubate the tool, which used a large language model to suggest code inside a developer's editor. GitHub previewed Copilot in 2021, and Warner has described it as an early demonstration that AI could work alongside programmers rather than merely autocomplete text.[1][3] His experience watching Copilot emerge from a general-purpose model later informed his conviction that a system built specifically around code could go much further.
After leaving GitHub, Warner joined the venture capital firm Redpoint Ventures in 2021 as a managing director on its early growth team. In that role he advised and invested in infrastructure and developer-tooling companies, including a backing of the blockchain infrastructure startup Alchemy.[2][3] He has said that the investing role gave him a broad view of the AI landscape and helped crystallize the thesis that became Poolside.
Warner and Eiso Kant, a Dutch software entrepreneur who had earlier founded the code-analysis companies source{d} and Athenian, co-founded Poolside in May 2023. The two had co-hosted a podcast on engineering leadership since 2020 and developed the idea for a frontier AI company focused entirely on code, software, and version-control history through their conversations there.[2][8] Warner became chief executive officer and Kant chief technology officer; the two are now generally described as co-founders and co-chief executives, with Kant continuing to lead the company's technical research.[9]
Poolside is headquartered in San Francisco and operates a substantial team in Paris and elsewhere. Although the Paris presence led some press to describe it as a French company, Kant has publicly clarified that Poolside is an American company with a globally distributed team, and that it declined sizable French research grants.[10] By late 2025 the company employed roughly 150 people.[8]
The company raised money quickly. It opened with a 26 million dollar seed round in May 2023, led by Redpoint Ventures, with partner Erica Brescia taking a board seat.[8] By the time of its next major raise, Poolside had taken in roughly 626 million dollars in total. In October 2024 it closed a 500 million dollar Series B led by Bain Capital Ventures at a valuation of about 3 billion dollars, with participation from Nvidia, eBay Ventures, DST Global, StepStone Group, Citi Ventures, Felicis, and Redpoint.[2][11][12]
| Date | Round | Amount | Valuation | Lead investor |
|---|---|---|---|---|
| May 2023 | Seed | 26 million dollars | not disclosed | Redpoint Ventures |
| October 2024 | Series B | 500 million dollars | about 3 billion dollars | Bain Capital Ventures |
| Announced October 2025 | Series C (attempted) | about 2 billion dollars target | about 12 billion dollars | Nvidia (up to 1 billion dollars) |
In December 2024 Poolside reached a strategic agreement with Amazon Web Services to offer its models on Amazon Bedrock and EC2, positioning the company to serve enterprise customers inside their own cloud environments.[13] Warner has steered Poolside toward regulated and security-conscious buyers, including government and defense customers, emphasizing deployment fully inside a customer's security boundary for classified, disconnected, or sovereign environments.[2][8]
In October 2025 Nvidia announced plans to invest up to 1 billion dollars in Poolside as part of a roughly 2 billion dollar financing round that would have valued the company at about 12 billion dollars, a fourfold increase over its 2024 valuation.[14] Around the same time, Poolside partnered with CoreWeave on Project Horizon, a planned 2 gigawatt data center campus in West Texas built around Nvidia GB300 systems and tens of thousands of GPUs.[15]
Those plans unraveled in early 2026. The roughly 2 billion dollar round, including the prospective Nvidia anchor investment, failed to close, and in early April 2026 CoreWeave terminated its anchor lease for Project Horizon, citing a shift toward more flexible, multi-tenant leasing.[16] Poolside began seeking new infrastructure partners for a scaled-down version of the campus. Reporting in April 2026 indicated that the company had paused most non-engineering hiring while maintaining a cash runway into late 2026, and that it had not conducted broad layoffs. Warner, who had stayed publicly quiet through the unwinding, posted on April 24, 2026 that "hard moments build hard companies" and that the team was "heads down on the model."[16] As of mid-2026, Warner remains co-chief executive of an independent Poolside, with the company's most recent completed raise being its 2024 Series B.
Warner has framed Poolside's goal as reaching a form of "narrow AGI through software and code," arguing that a model specialized for programming could surpass general-purpose large language models on engineering tasks. Rather than treat coding as one capability among many, the company concentrates its data, training, and infrastructure entirely on software development.[2][3]
The central technical idea is Reinforcement Learning from Code Execution Feedback, abbreviated RLCEF. Most language models learn to write code by reading large quantities of existing source code. Poolside instead has its models attempt realistic coding tasks against real code bases, then run the resulting programs, execute tests, and inspect compilation errors, using the outcomes as a reward signal in a large-scale reinforcement learning loop. The intent is to mimic how human engineers actually improve, by writing programs and learning from whether they run correctly rather than only by imitation.[2][17] Generating this execution feedback at scale requires heavy infrastructure: Poolside has described running on the order of a million container images and roughly 10,000 code executions per minute to produce synthetic training data.[17]
Poolside packages this work in an internal training system it calls the Model Factory, an industrialized pipeline for building and evaluating its foundation models. The company has released models including Malibu, a versatile model aimed at complex engineering tasks such as multi-step code, test, and documentation generation, and Point, a smaller model tuned for fast, context-aware code completion.[17] Consistent with Warner's enterprise focus, these models are offered for deployment within a customer's own infrastructure, so that proprietary code does not leave the organization, an approach intended to appeal to large enterprises and public-sector users with strict data-control requirements.[2][13]