- Joined Cognition
- Jul 2025
- How they joined
- With Windsurf
- Before Windsurf
- Workplay LabsEntrepreneur in Residence
- Status
- Current
- Jeff Wang was Windsurf's Head of Business before stepping into the interim CEO role effective July 11, 2025, after Varun Mohan, Douglas Chen, and some R&D team members left to join Google.Jul 11, 2025
- He negotiated the Cognition acquisition over a single weekend, closing within roughly 72 hours: a first call after 5 p.m. Friday, a signed definitive agreement by Monday morning, and an announcement that afternoon.Jul 2025
- He is now President of New Enterprise at Cognition, the company behind Devin, described as the world's first cloud agent.
- In 2026 he was publicly launching and promoting Cognition products including Devin Desktop, Devin Oncall, Devin Security 2.0, Devin Fusion, FrontierCode 2.0, and Memory and Dreaming.Oct 23, 2026We’ve brought everything great about Devin Cloud into Devin Desktop. It’s been reconstructed and designed with having the best of both worlds of local and cloud agents Try it out today! We suspect that the Desktop app will be the default app you run your agents on
- He joined Codeium (later renamed Windsurf) in May 2023.2023
- He earned a B.S. in Electrical Engineering from the University of Michigan in 2008 and an MBA from UC Berkeley's Haas School of Business in 2020.2020
- He spent roughly two years building Windsurf's entire go-to-market engine from scratch as Head of Business, which he described as 'nothing short of world class.'
- His first deal was $1,000, scaling to seven-figure contracts three to four months later, and he helped launch Codeium's paid cloud product and self-hosted AI platform, signing Dell, Anduril, and Clearwater.
- He welcomed Corridor to the Cognition team in October 2026.Oct 29, 2026@cognition @walden_yan @corridor Welcome to the team!
- He has emphasized security repeatedly in 2026, saying companies are not taking security seriously enough and that 'we're entering a new era,' while shipping Devin Security 2.0 and open-sourcing an agent plugin to patch critical PHP, Node.js, and GitLab flaws.Oct 26, 2026We've entered a period of uncertainty where there will be scary security news all the time, and worse yet, the capabilities of bad actors will continue to get more threatening So at Cognition we're doing a few things. First we're releasing Devin Security 2.0. We've made it easier to find vulnerabilities, validate, prioritize, and fix. This also checks the rest of your software stack for dependencies and gives you a comprehensive solution, it isn't a simple security scanner. You'll get the full power of Devin and models like Mythos We're also giving millions of dollars of credits away for our
- He helped take Codeium to a $500M valuation within a year, following a $65M fundraise led by Kleiner Perkins and General Catalyst announced in early February 2024.Feb 1, 2024
- He had been part of Windsurf's journey since mid-2023, serving across many roles and capacities as the team expanded.Jun 2023
Created an education platform for cryptocurrencies, economics, AI, and VC Host of RocketFuel Crypto and Rich Dad Crypto, a weekly video and monthly newsletter of all things crypto
Ran product for Tomo to transition from a Series A to B company Set Web3 custodian, infrastructure, and roadmap for Tomo's platform Created B2B and partnerships structure Seed round investor
Great seeing Devin used to improve inference efficiency, this is something that normally takes many engineers but using Devin to keep looping for improvements is likely how things will be done going forward https://lnkd.in/gw-K5FRB
Great seeing Devin used to improve inference efficiency, this is something that normally takes many engineers but using Devin to keep looping for improvements is likely how things will be done going forward
Every so often models catch up to benchmarks and it gets difficult to discern them from each other. At Cognition we’ve released our new eval FrontierCode 2.0, it’s more thorough and accurate than 1.1, and really showcases the gap between models. We also use this internally on how to choose models for tasks and getting the best bang for buck on running Devin Check out the leaderboard today and we’ll continue tracking where models stand! https://lnkd.in/gkmSjNN6
Every so often models catch up to benchmarks and it gets difficult to discern them from each other. At Cognition we’ve released our new eval FrontierCode 2.0, it’s more thorough and accurate than 1.1, and really showcases the gap between models. We also use this internally on how to choose models for tasks and getting the best bang for buck on running Devin Check out the leaderboard today and we’ll continue tracking where models stand!
We've entered a period of uncertainty where there will be scary security events all the time, and worse yet, the capabilities of bad actors will continue to get more threatening. So at Cognition we're doing a few things. First we're releasing Devin Security 2.0. We've made it easier to find vulnerabilities, validate, prioritize, and fix. This also checks the rest of your software stack for dependencies and gives you a comprehensive solution, it isn't a simple security scanner. You'll get the full power of Devin and models like Mythos We're also giving millions of dollars of credits away for
We've entered a period of uncertainty where there will be scary security news all the time, and worse yet, the capabilities of bad actors will continue to get more threatening So at Cognition we're doing a few things. First we're releasing Devin Security 2.0. We've made it easier to find vulnerabilities, validate, prioritize, and fix. This also checks the rest of your software stack for dependencies and gives you a comprehensive solution, it isn't a simple security scanner. You'll get the full power of Devin and models like Mythos We're also giving millions of dollars of credits away for our
We’ve brought everything great about Devin Cloud into Devin Desktop. It’s been reconstructed and designed with having the best of both worlds. Get very snappy and fast performance locally while still getting access to all your workloads running in the cloud Try it out today! We suspect that the Desktop app will be the default app you run your agents on
We’ve brought everything great about Devin Cloud into Devin Desktop. It’s been reconstructed and designed with having the best of both worlds of local and cloud agents Try it out today! We suspect that the Desktop app will be the default app you run your agents on
Today was one of the biggest days in cybersecurity reporting and incidents with PHP, node.js, Gitlab, and more As such, we’ve open sourced an agent plugin that can be used to help patch these. These can also be used in Devin We’re also giving significant credits to our Enterprise users so that they can focus on remediating these. Devin is uniquely effective here since it can validate that other parts of your applications are still working after the changes Going forward stay safe, this problem won’t go away but you can put things in place to defend against it Please see here for more inf
Today was one of the biggest days in cybersecurity reporting and incidents with PHP, node.js, Gitlab, and more As such, we’ve open sourced an agent plugin that can be used to help patch these. These can also be used in Devin We’re also giving significant credits to our Enterprise users so that they can focus on remediating these. Devin is uniquely effective here since it can validate that other parts of your applications are still working after the changes Going forward stay safe, this problem won’t go away but you can put things in place to defend against it
Wow, Devin Fusion just got the top score of 68 on Artificial Analysis's Coding Agent Index (using a mixture of Opus 5.5 and GPT-6 Luna) while being 31% cheaper than second place There are actually a lot of reasons why we're able to get the top score for cheaper There's simply no single model that is the best at all tasks for the price. By figuring out what tasks different models are good at, we could be beating certain tasks even from cheaper models vs. frontier models. However this does come at the cost of swapping models and refreshing the prompt cache This is why Devin Fusion is awes
There are actually a lot of reasons why we're able to get the top score for cheaper There's simply no single model that is the best at all tasks for the price. By figuring out what tasks different models are good at, we could be beating certain tasks even from cheaper models vs. frontier models. However this does come at the cost of swapping models and refreshing the prompt cache This is why Devin Fusion is awesome, it tries to maintain the cache as long as possible while working in parallel with cheaper models. It's able to also verify and fix the final output with frontier models, many rou
