What happened to Devin from Cognition AI?

On March 12th, 2024 a company called Cognition AI released a bunch of videos, including the one titled "Introducing Devin, the first AI software engineer" on YouTube. They claimed to show a working demo of the autonomous AI engineer called Devin. The idea was revolutionary and novel. An AI engineer writes code, executes it, and fixes the bugs it faced on its own. The video demonstrates how the AI Engineer writes up a plan, and starts executing and drives through bugs. Devin claimed to be have its own integrated command line, code editor and browser. The video claims further that Devin can fetch and read API documentation and starts building based off of that. 

The video went viral on the Social media and it actually frightened developers across the world (at least some of them). There were mixed reactions from people. After some time, people started scrutinizing the demo video published by Cognition AI. A YouTube channel called "Internet Of Bugs" debunked one of those videos and exposed the lies spread by Cognition AI across the social platforms like YT, X etc. Interesting video, you can watch it here if you haven't watched already- "Debunking Devin: "First AI Software Engineer" Upwork lie exposed!"

The whole debunking and fear of loosing jobs caused Cognition AI to face a lot of backlash and hate from the developer community. Everybody was talking about Devin. 

Believe me, despite facing a big backlash from the people, Devin actually fueled the idea of agentic coding. Devin actually worked as a catalyst to turn the Agentic Coding from a niche experiment to massive industrial shift and commercial value.

Despite its faults, Cognition AI actually proved that the Agentic Workflows could solve real problems. It pivoted the major players from text-based autocompletion to agentic coding. Even with bad marketing and fabricated demos, they were able to prove the value to the companies. The main problem of the marketing was that the targeting was bad. They tried to target individual developers instead of focusing the enterprise  clients. Selling expensive autonomous agents to developers was very a bad position. And they failed miserably on that. People (developers) didn't like the idea. 

But, it was not a total failure, the video, created an instant fomo among the top tier VC's. As they were already tired with the copilot and text-autocompletion, this demonstrated the value for VC's.

That worked very well for them and they were able to secure multiple rounds of fundings. Here is the sequence of the funding they received.

Before going out public on YT.

• Pre-Launch Stealth Funding (Seed) - they used this funding to build the prototype.

Right after the video went viral, they secured funding as follows:

• March 2024: Public Series A, $21 Million
• April 2024: $175 Million, the initial hype was still not wore off until this time. Surprisingly, this was the time when the fake demo controversy happened (video mentioned above).

Failed individual plan rollout (initially they had targeted the individual developers):

• Late 2024: people called it expensive toy and didn't serve the expectations of the individual developers.

This was the low time for the company as the things didn't work out and the year round turnover till Sep 2024 was only around $1Million. 

In March 2025, despite of public skepticism, Cognition AI secured $4B valuation in another funding round. Then the important shift came into the future of Cognition AI, instead of focusing on individual developers, they started focusing on the enterprise centric solutions. They released Devin Wiki, which could read massive enterprise documentations. Companies like Goldman Sachs, Nubank, Evinova had already started using Devin Wiki.

Acquisition that mattered a lot: Around July 2025 google hired the top leadership of the Windsurf AI, which created void in the Windsurf, and the rest of the company was up for acquisition. In the right moment Cognition AI acquired the Windsurf AI. This allowed Cognition to integrate the isolated AI agents with the Windsurf's popular IDE. 

What did they actually deliver to their customers?

Traditionally SAAS charges seat bases prices, eg $20/Month. Basically a fixed value per user seat. Cognition completely bypassed that limitation by charging based on Agent Compute Units (ACUs). 

When an enterprise bank deploys an "Army of Devins" to run hundreds of automated code-patching tasks concurrently in the background, they consume massive amounts of ACUs. A single corporate client can easily scale their usage to millions of dollars per month based on the sheer volume of work executed.

Basically hundreds or thousands of agents started patching the old, boring and outdated code-based for the enterprise banks. Believe me this is very hard to maintain with Human engineers. Nobody likes this type of works. 

How did the convince the enterprise customers?

To close the multi-mill $ deals, they used a $10M productivity guarantee. They built a proprietary tool which could audit and translate the work done by Devin into actual engineering hours. Basically time required for human engineers to fix those issues manually.

The contract was like this, if Devin fails to provide value, Cognition bear cost up to $10M. This completely eliminated the financial risk for corporate buyers, allowing enterprise sales teams to close massive deals entirely behind closed doors.

Conclusion

The story of Devin tells us about a very different reality of how the enterprise Software sells. It proves that the initial hype or shame of the company doesn't matter in the enterprise world. This company actually traded the LinkedIn noise for the enterprise contracts and moved from the fake demo to enterprise company with ARR of $492M (Till May 2026) and $47B valuation. the hype died, shame vanished but company survived.