A year ago this month, Cognition acquired Windsurf in what its own founders now call “the craziest 72 hours of our lives.” The anniversary note from Scott Wu and Jeff Wang, published July 14, fills in the shape of the deal and the shape of the year that followed — and the numbers in it are large enough to be worth slowing down for.
The deal story is already well-trodden, including in CNBC’s reporting at the time: a first call after 5 p.m. on a Friday, a printed letter of intent carried to Windsurf’s offices on Saturday, a signed definitive agreement by Monday morning. What the anniversary post adds is the year of output that followed, and a few figures that put the acquisition in different perspective than the deal-price headlines did.
The receipts
Cognition’s own accounting of the year, drawn from the anniversary post:
The product record is the part that can be checked against shipped software. Three days after the deal closed, Cognition shipped Wave 11 — titled “Just Keep Shipping” — and the anniversary post frames it as the tone-setter for the year. What followed, by Cognition’s own list:
- Three homegrown models: SWE-1.5 (served at ~1,000 tokens per second), SWE-1.6, then SWE-1.7, described as the most capable and efficient model Cognition has trained to date.
- New product surfaces: Devin Review, Devin CLI, Devin 2.2, and Cognition for Government.
- A brand unification: Windsurf 2.0 consolidated agents into a single command center, then the whole thing was rebranded as Devin Desktop — “your home for software engineering.”
- An evaluation bet: Arena Mode and a public leaderboard, paired with the “Switzerland of AI” positioning on model choice.
- A commercial guarantee: the AI Productivity Guarantee, and new partnerships with automakers and systems integrators.
- A research line: FrontierCode and Devin Fusion in June 2026, then Devin Security Swarm and a Vulnerability Remediation Program in July.
The monthly ship log in the post runs from July 2025 (Voice Mode, Named Checkpoints, MCP Marketplace) through July 2026 (Security Swarm) without a blank month.
The acquisition thesis, a year later
The framing Cognition offers for why the deal worked is structural: “Cognition had built one of the best engineering orgs in Silicon Valley and was looking for a GTM org; Windsurf had built one of the best GTM orgs and was looking for engineering. Our products didn’t even overlap: one of us had been building a cloud agent, the other an IDE.”
A year of shipped product is consistent with that thesis. The IDE company became the command center for the cloud agent company. Devin Desktop’s Agent Command Center — documented as a Kanban that puts local Cascade/Local sessions beside cloud Devin VMs — is the literal productization of “one boat.” The CLI’s /handoff to cloud Devin is the seam between the two original products, now a single workflow.
The community read
The anniversary drew community posts that tracked the same shape. The most shared framing, from @the_vc_intern, compressed the year into the same bullet list Cognition’s post leads with: Wave 11, SWE-1.7, Devin Review, Devin CLI, unified Devin Desktop. A separate thread from @sandeepDshah pointed at the part the deal-price coverage underweighted: “Most of the conversation centered around the founders & engineers heading to Google. I knew the bigger story was about who stayed.”
That is the read the anniversary post itself tacitly endorses — it is signed by Scott Wu (Cognition) and Jeff Wang (Windsurf side), and the ship log is the answer to “who stayed, and what did they build.”
What to actually take from it
If you are evaluating Devin as a team, the anniversary post is a useful inventory of which surfaces are mature and which are new. The product taxonomy is now stable enough to plan against: cloud Devin for async, long, or parallel work; Cascade for the in-editor pair partner; Devin Local / CLI for the preview-path local harness that is meant to eventually replace Cascade. The model line is its own decision — SWE-1.7 is free on Pro and above, and Cognition’s Arena Mode positions model choice as a user decision rather than a vendor lock-in.
If you are watching the company, the number worth tracking is not the deal price or the headcount. It is whether the $500M+ run-rate claim converts into retained revenue after the AI Productivity Guarantee is exercised at scale — and whether the unified Desktop holds the IDE audience that made Windsurf worth acquiring. A year in, both are still forward-looking questions. The shipping cadence is not.
References
- One Year of Building Together — Cognition blog, Scott Wu & Jeff Wang, 2026-07-14
- CNBC: Cognition to buy AI startup Windsurf days after Google poached its CEO — CNBC
- Introducing Devin — docs.devin.ai
- Devin in Devin Desktop — docs.devin.ai
- Agent Command Center — docs.devin.ai
- SWE-1.7 launch — Cognition blog
- Devin Fusion — Cognition blog
- FrontierCode — Cognition blog
- @the_vc_intern — acquisition anniversary thread — X
- @sandeepDshah — on who stayed — X
From Devin Central — subscribe for Devin news, features & tutorials. Devin is a trademark of Cognition; other product names belong to their respective owners.