Dan
Adler

The CEO who never stopped shipping

There is a version of this story where Dan Adler walks into Sourcegraph's demo in fall 2016, watches the cursor jump from a function call to its definition across two separate repositories without leaving the browser, and has the kind of experience that later gets polished into a founder's origin myth. He doesn't polish it. He just said, out loud: "How does every developer in the world not have this?" - and then joined the company.

That bluntness is structural. Adler grew up as a programmer, studied computer science at Rice University (class of 2007 - before the iPhone existed), spent four years as a Bain consultant, two years doing private equity investing, then got an MBA at Stanford where he was named an Arjay Miller Scholar. By the time he walked into Sourcegraph, he had spent a decade inside the machinery of business. What startled him wasn't the idea. It was that the gap was so obvious, and so persistently unaddressed.

Nine years later, on December 2, 2025, he became Sourcegraph's CEO - elevated when co-founder Quinn Slack departed to lead Amp Inc., the AI coding agent spinout. The announcement was quiet by tech-industry standards. No long Twitter thread. No breathless press cycle. Just a note on the Sourcegraph blog and a company that had already been running on the infrastructure Adler himself built.

What makes Adler unusual is the refusal to split. Most operators go one way: they run the business or they write the code. Adler refused the binary. He wrote Sourcegraph's original data infrastructure. He sold most of their first contracts and gave their first demos. He built their first dashboards and supported their first enterprise customers. Then he became VP of Business, CFO, and infrastructure engineer - sometimes all at once. git blame on Sourcegraph's codebase shows his name scattered across nine years of files, a topographic map of someone who never fully left the engine room even while managing everything above it.

What Sourcegraph actually is

Sourcegraph is not a code editor. It is not a linter. It is not a test runner. It is the layer that sits underneath all of those things and answers the hardest question in enterprise software engineering: where is everything, and how does it connect?

Codebases at scale are not complicated - they are incomprehensible. A company like Uber or Stripe or Atlassian doesn't have a codebase. It has thousands of interlocking codebases, spread across repositories, built in multiple languages, modified by hundreds of engineers over years. Sourcegraph indexes all of it - 54 billion lines of code across paying customers - and lets developers search it, navigate it, automate across it, and now, run AI agents over it.

The Series D closed in July 2021 at a $2.6 billion valuation - $125 million from Andreessen Horowitz, with Insight Partners, Geodesic Capital, and others participating. ARR hit $10 million by the end of that year. By 2025 it was $50 million. Adler had a hand in building every system that made those numbers possible.

Why code context matters more now, not less

The argument that Adler inherited when he became CEO is a counter-intuitive one: that the rise of AI coding assistants makes Sourcegraph more valuable, not redundant. The conventional take is that LLMs replace search. Adler's take is that agents are blind without indexing.

An AI coding agent operating on a large enterprise codebase without proper context retrieval isn't just inefficient - it hallucinates. It recommends importing packages that don't exist in that codebase. It misses the dependency graph. It cannot see that a function called by this microservice is being deprecated by a different team on a different branch. Sourcegraph's "Deep Search" - its agentic retrieval layer - is built on the thesis that RAG hasn't declined in importance; it has become load-bearing infrastructure for anyone building agents at enterprise scale.

When Sourcegraph and Amp separated in December 2025, the logic was precisely this: Amp is a consumer-facing AI coding product that competes for the attention of individual developers. Sourcegraph is the infrastructure layer that enterprises require regardless of which AI product their developers prefer. Different distribution models. Different buyer psychology. Different everything. Splitting them was clean strategy, and the board - including Slack and Beyang Liu, the two co-founders - remained on Sourcegraph's side of the table.

Nine years in, one decade before

What got built along the way

• ■ Helped grow Sourcegraph from zero revenue to $50M ARR and a $2.6B valuation
• ■ Wrote Sourcegraph's original data infrastructure as an early employee
• ■ Sold most of Sourcegraph's first enterprise contracts and ran early customer support
• ■ Accumulated 100+ production commits across nine years on the platform
• ■ Served simultaneously as CFO and infrastructure engineer - a combination unique to early Sourcegraph
• ■ Built Sourcegraph's people ops, legal, finance, business ops, and tech ops functions from scratch
• ■ Co-inventor on U.S. Patent No. 9,946,702 (2018) for a multicomputer data transfer system
• ■ Honored as Arjay Miller Scholar at Stanford GSB (top 10% of graduating class)
• ■ Elevated to CEO in December 2025 with the full confidence of the founding board

The texture of the operator

There is something revealing about Adler's GitHub account. His username is dadlerj. His repositories include a Go project called tin (64 stars - a modest but respectable number for personal infrastructure work), a Ruby XLSX reader called dullard, and a Go documentation generator called gowalker. He has earned the "Pull Shark" badge four times - meaning at least 16 pull requests merged from others. He has the "Arctic Code Vault Contributor" achievement, which means some of his code is preserved in a mountain in Norway for a thousand years.

The GitHub achievements matter because they are evidence of the thing that's hard to fake: sustained technical participation, over years, in open codebases, across languages. Not a "technical CEO" who coded once and stopped. Someone who kept going.

His first startup, Cells Software, didn't make it. The product was focused on labeling data inside spreadsheets - an idea that was adjacent to the machine learning toolchain conversation that was just beginning to heat up in 2015. He wound it down. This is not mentioned as failure. It is mentioned as evidence of pattern: Adler identifies underserved infrastructure problems - how spreadsheet data gets labeled, how code gets searched at enterprise scale - and then tries to fix them.

He lives in Oakland with his fiancee and dog. He joined Twitter in November 2011. His posts are infrequent.