←
Jeff Wang2,387 posts
Jeff Wang avatar

Jeff Wang

@jeffwang

President, New Enterprise at Cognition, Prev. Windsurf CEO, AI @ jeffwang.substack.com

◷ Joined February 2009▧ 272 Photos and videos
Jeff Wang avatar
Jeff Wang@jeffwang·20h

Today is our biggest launch of 2026 Every time models get better, I’ve been lucky to witness the evolution of AI Coding: Autocomplete/Chat at Codeium Local Agents at Windsurf Cloud Agents at Devin Now we’ve launched Persistent Agents with Devin Bots Before, you would tell an agent a specific task and be very clear about the expected outcome, but interacting with agents has changed again, you can now give high level objectives to an always on agent and it will act like a team mate to complete them and check back with you. It’s also always there to check in, preserving memory and context, as well as working with other Bots and humans as well, just like a good teammate

Meet Devin Bots. From writing firmware to merging thousands of PRs a week, a Devin Bot leads a team of agents to get the job done.

1:33
Jeff Wang avatar
Jeff Wang@jeffwang·4h

This is what it looks like to have a team of Devin Bots managing an entirely new product with very little human intervention (and only 1 human) Devin built our Android app from scratch, with only our iOS app as reference. The Bots showed its work, got feedback when needed, and tested itself to make sure it was polished - 88% of its PRs were self-merged

I was future-pilled on Devin Bots after seeing us build and maintain our Android app with 0 human engineers. It's owned by an autonomous swarm of Devins which are goaled on maintained feature parity with iOS. @samyok who created the Bot has never built an Android app or written Kotlin.

1:23
Jeff Wang avatar
Jeff Wang@jeffwang·20h

I can’t tell you how much easier it’s been to keep driving and keeping track of things happening across our company, but this doesn’t come without caveats. We want to educate our users with how to best use Bots, as well as how to allocate budgets for the use cases that make the most sense. Internally, this has been the most addictive and productive product in the shortest amount of time, we can't wait to hear from you

Jeff Wang avatar
Jeff Wang@jeffwang·20h

Use a Bot to continuously improve your app performance, continually improving infra, or maybe even “keep my calendar free from conflicts”. I’ll show you a few things I’ve built for myself at Cognition: Triage Bot: Which continuously scrapes through Slack, Notion, Salesforce, Granola notes. This bot then feeds other Bots relevant information New Enterprise Bot(s): I have multiple bots that acts as a PMs for new products, keeping track when customers ask for specific things we don’t have, or when we have design partnerships in the field Enabler Bot: This bot keeps all the New Enterprise Bots accountable for having a pitch deck, qualifications, and a log of all activity Piper: My sales pipeline bot, which tells me if a deal does not match the customer notes, it’s literally AGI when it proactively pings me on hearing something in the notes not matching what is forecasted or entered correctly in Salesforce JeffCal: My EA who surprisingly does a good job of prioritizing between other Execs

Post attachment
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 13

Vibe coding for enterprise never quite took off. Who wants to figure out all the security and hosting complexities? But what if we took care of this using Devin’s compute stack? We’d make it so apps could only be accessed by your org, and access controls to data would be shared via the access Devin already has So introducing Devin Apps, we made sharing internal tools and apps viable, and spin down apps when they are not in use

Introducing: Devin Apps Private, secure, sharable web applications running directly on Devin’s cloud computer. Devin Apps are designed to power your most ambitious internal tools, preview deployments, and extensions to Devin itself.

1:58
Cognition avatar
⟳ Jeff Wang retweeted
Cognition@cognition·Nov 10

Claude Mythos 5.2 is now available in Devin’s managed security scans. We are giving $100M+ in credits to help companies find and fix their most critical vulnerabilities with Claude Mythos 5.2. Eligibility criteria and conditions apply. On our internal cyber benchmark, Mythos 5.2 scores 73.5, a substantial step up from Mythos 5.1’s 56.9.

Post attachment
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 5

We've made Devin Review 5x cheaper, it'll cost you less than a dollar to review a PR now, and we made it much faster as well

Introducing Devin Review 2.0: the best way to review PRs with Devin. Now up to 5× cheaper, 1.5× faster, and with new features and integrations.

0:40
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 5

With how comprehensive these models are getting in terms of finding cybersecurity vulns and unpredictable behavior, having an environment as close to real life will start to become more and more important for agents to test against

introducing simulate.ai by @raindrop_ai test your agent on every PR, against real scenarios, with one-click. we handle all the set up: sandboxing, fixtures, real data, synthetic data, evals and more.

2:07
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 4

We've released our latest model SWE-2 Flash based on Inkling-Small from @thinkymachines By post-training the model, we were able to outperform all American open-weight models as well as get the cost to be 20% of SWE-2 while making it 3x faster We're going to make this free for the rest of November, so give it a try today!

Introducing SWE-2 Flash, our new fast & efficient coding model. On leading evals, it outperforms all American open-weight models, even ones 3x larger. To balance transparency and cyber risks, we are building a program to securely share the weights with partners & researchers.

Post attachment
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 4

Congrats @gauri__gupta on the launch!

The intelligence frontier keeps advancing every day. Are your agents delivering better outcomes on what matters? For enterprises, the promise of agents is more work completed reliably, with less human intervention and a lower cost per successful task. Today, we’re launching @neosigma_ai and our mission to empower every organization to build their own sovereign intelligence lab. A place to operationalize their knowledge, replay real work, and optimize agents against the outcomes that matter to their business. Where production experience compounds into intelligence the organization owns. We’re starting with NeoEval, our automated evaluation engine. It turns your production data into evaluation environments - that dynamically evolves along with your users, workflows, and frontier. A playground for evaluating and optimizing your agents. Your data. Your institutional knowledge. Now operationalized into evals and compounding intelligence. Stay tuned for the next few days of updates from NeoSigma.

1:51
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 3

Right now, I do not think companies are taking Security as seriously as they should. Here are a couple of best practices to follow from @jackhcable

It's been a rough few weeks for security. I pulled together a guide of the highest-impact actions you can take today:

nader dabit avatar
⟳ Jeff Wang retweeted
nader dabit@dabit3·Nov 3

SWE-2 is now free in Devin (CLI and Desktop) until November 30 for all subscriptions ⚡

0:10
Jeff Wang avatar
Jeff Wang@jeffwang·Nov 2

Congrats on the Trajectory launch! Does this mean that I too, can be an AI engineer

Imagine if everyone had the tools to shape intelligence itself. That day is today. Today we launch the Trajectory platform - the easiest and most intuitive tool to post-train models of your own. The foundation for continual learning. With just a few agent prompts, you get access to a full creative suite to train, eval, and deploy trillion parameter models. On top of that, we believe in a future where you don’t just shape your model once, it also keeps learning on its own from the world around it. We’re releasing two research previews. 1. Continual grading scales your taste, so your model learns what good looks like from far more than you could ever grade by hand. 2. GS1, a new algorithm that lets models learn directly from what they do in production. We’ve now raised 60M, with a new round led by @Sequoia at a 300M valuation to democratize the most powerful technology of today - training models - so that the lawyers, doctors, and engineers of the world can own their own intelligence. You know what good looks like, now go build it. Own your own value function with Trajectory.

1:43