@ryancarsonOct 27

Had a nice dinner with @imjaredz and the @DevinAI @cognition crew last night. Couldn't live without Devin at @HelloUntangle. Switch to a great cloud harness. If you're doing local dev, you're working in the past. Try it and you'll see the light.

@mtsliveOct 27

Cognition CSO @alexstamos warns we're entering a "dark age of cybersecurity" as AI exposes decades of software vulnerabilities faster than companies can fix them: "We came out of the Dark Ages. We had the Enlightenment, we had the Renaissance. This will end, just like the real Dark Ages ended." "We have decades of code that was written by humans, often in non-type safe languages, that cannot hold up against these superhuman bug-finding AI systems. We are slowly burning through all of the bugs in that software." "The PHP bug was discovered by researchers who got free credits from OpenAI. Totally the right thing for OpenAI to do, but the moment they’re released, attackers now have the ability to much more quickly write exploit code." "There’s a lot of code that should just be rewritten from scratch instead of fixing thousands and thousands of bugs, rewritten by AI with human oversight in languages that we know are more secure." @cognition

@dabit3Oct 26

Known affiliation

The world's most powerful cyber models, including Claude Mythos, are available in Devin Security Platform. We're also offering hundreds of millions of dollars in credits to use it. If you want to defend faster than attackers, you have to find and fix flaws at machine speed. That’s why we’re building the best AI-powered security platform in the world. Devin already helps engineering teams write more secure code, now it can secure the code they already have.

@svpinoOct 20

In 1994, software was dead. People told me programming was for nerds and that I should become a doctor or a lawyer instead. In 2004, software was dead. People told me that someone from India would take my job for $5/hour. In 2014, software was dead. People told me that no-code was going to kill every job and doom my career. In 2024, software was dead with Devin, Windsurf, Copilot, and OpenAI o1. We kill software every 10 years, and yet the demand for good Software Engineers has never been higher.

@dabit3Oct 19

Known affiliation

Devin Fusion tied #1 on Artificial Analysis Coding Agent Index Outperforms Claude Code w/ Opus 5.5 on both performance and price, at 31% lower cost per task Fusion has a wide array of model combinations, so as more models become available the design space for finding optimal / specialized combinations grows.

@jaredpalmerOct 19

One-size-fits-all blanket usage caps on AI spend don’t make sense. On the other hand, yolo tokenmaxxing doesn’t scale either. We’ve come up with a better approach to cost control centered around efficiency. More efficient employees can have limits raised automatically (with a final upper bound still), while other spend requests go thru an admin review process.

@dabit3Oct 16

Known affiliation

Devin can now build and test iPhone Duo apps in macOS cloud VMs Unfold the phone, fold it, close it, check every screen, and get video verification of everything. Here are some examples: 1. Devin takes someone's portrait while they watch themselves on the outside screen, it sends posing cues, half-folds the phone so it stands on a table, takes timer shots, and closes the phone mid-countdown to make sure no photo gets lost

@jaredpalmerOct 15

We productized how we handle incidents, bug reports, and alerts at @cognition with Devin Oncall. It’s completely changed how we work and also how we think about incident management and SRE. From investigations, to reproductions, to triage, rollout, root cause analysis and even postmortems, we do it all with Devin. Excited to share this with the world.

@svpinoOct 13

I've been using Kylon for a few days already, and I really like it. This is a "company harness". It takes your people + your data + your agents and puts them into the same workspace. Basically, everyone can work and collaborate in the same space. I'm using it with my local Codex. You can also connect it to Claude Code, Devin, or use a Kylon Cloud account. When you run it locally, Kylon hands off any agentic task to my Codex. It works pretty well! If I had to guess, most companies will move eventually to a model like this. It just makes sense.

Dynamic workflows let Devin split a big job into pieces, hand each one to its own Devin, and bring the results back together. The plan for a Dynamic Workflow lives in code instead of an agent’s head, so you always know which Devins will run, in what order, and what each one is told.

Devins can start managed Devins of their own, so one task can branch into a tree of Devins several levels deep. Work done in sequence takes as long as all its parts added together, while work split this way takes roughly as long as its slowest part. So size stops being the reason a task gets postponed, since a larger task simply produces a larger tree.

@dabit3Oct 8

Known affiliation

so many design debates existed because building used to be expensive. at @cognition instead of debating an idea, we build competing ideas in parallel and debate working software to take this a step further, you can just build directions up front and skip some of the original design discussion, bring results as the first point of discussion Devin cloud agents build multiple directions in parallel, send screenshots / previews / videos to review in Slack along with PRs. when we agree on an implementation, it's ready to merge. velocity goes way up!

@ryancarsonOct 8

If your company or startup is running engineering with everyone using their own license of Codex/CC then you're going to be completely left behind by fast moving teams who are using cloud-based agents. Local dev is like banging rocks together and grunting versus being on a rocket. Seriously, just pay the money for something like @DevinAI, @AmpCode, @FactoryAI, etc. Do NOT try to build your own "code factory". Props to @DaleSpoonemore for taking @HelloUntangle to the next level on this. Fun to watch.

@catalinmpitOct 7

T3 Code first impressions I'm writing everything that comes through my mind. Some may be non-significant or simply my skill issues, but I'll keep them because they reflect my experience. > Smooth and fast onboarding This is extra, but I wish I could select more accounts to connect (e.g. Devin) in this step. It only allowed me to connect Claude and OpenAI subs. Other than that, great. > "Add provider" option easy to miss? It'd be nice for the "Add provider" option to be underneath the existing connected providers. But maybe it's not the best option if you have many providers connected and need to scroll down to find the option. Or, at the very least, move the "+" sign next to the "Providers" text. (see the first image attached. I can't embed the picture here for some reason like I used to in the past) > UI mismatch for different projects Maybe I'm tripping. But the left sidebar shows the same project name for both tasks. The other views & terminal show the correct project names. I'm stuck on this one. Came back to it after writing the post and using T3 Code for longer. I asked T3 and apparently they're treated this way because it groups them by the repository path even though they are different folders. I have documenso, documenso-1 and documenso-2. Even though they're different folders, they point to the same repo. I'd like to be able to group the tasks by project - e.g. all documenso-1 tasks should sit in that project. After a lot of random commands, I made it work somehow. (see the second image attached) > Really cool UI The app looks really beautiful all around. Feels like it was very well-thought. Besides that, you can choose between themes, which gave me a throwback to the times when I used to manually code in VS Code and play around with themes. > It's fast I give it a task and by the time I try to do something else, the task is already finished. Annoyingly good. > Audio ping I'm used to Herdr pinging when a task is done. Would be nice to be able to do that in T3 too. Scratch that. After writing this, I Googled if T3 Code has the functionality because I couldn't find it. Had to read a GitHub issue to find the option. > PR page A nice PR page that shows your PRs and the ones that requested a review from you. Really useful since I work in open source and can go over PRs without changing context (go back and forth between GitHub and work app). > Task organization Would be nice if I could have the tasks split by "Working", "Done", "Review", etc. > Mobile app error I have the nightly version of T3 Code and couldn't connect the mobile app I downloaded from the App store. But I asked T3 what the problem is and it gave me the TestFlight link for the mobile version that should work. I followed the instructions and were able to pair the mobile app with my desktop instance. > Mobile app good The mobile app is simple and to the point. No unnecessary features. I like that you have access to the terminal and filesystem files. I rarely use apps for remote coding, but I could see myself using this one. > The end That's it. These are my first impressions after using the app to work on features on a real app with customers (@documenso). My only nitpicks are the project grouping and the UI choice for adding new providers. I'd like to have more options for grouping the projects and their tasks. Also, I'd move the "+" sign for adding providers from the right side. Other than that, the app is really good and there're more positive things than what I've highlighted in this post, but I don't want people to call me a shill. Will I use it going forward? If I reach for a desktop app for managing all my AI subs, probably. But for now I'm hooked on the terminal + herdr combo.

@jaredpalmerOct 6

I’ve been experimenting with Devin and autoresearch every night for the past week or so. The prompt I’ve been using for is pretty simple: `I want you to do autoresearch to improve {verifiable metric(s)}. Hill climb. You have a budget of ${budget} and {time limit}. Use sub-devins to parallelize work. Track your plans and progress as you go in the filesystem. Work methodically. Create abstractions if you need to to help you verify faster. Manage your own PRs/babysit code review and merge as you see fit until you get a candidate that passes gates you set. Boil the ocean. Believe in yourself. You got this. {+more confidence/hype statements}` After you have the setup you like, just make this into a skill and turn on automatic code review. With this prompt, Devin can run for hours autonomously. Funny enough, the “boil the ocean” and hype statements seem to yield better exploration behavior and make most models less likely to give up early or reward hack changes/ship incremental fallback stuff. Inspiration for this was @jarredsumner’s prompts during his Riemann work.

@ryancarsonOct 6

Just told a Devin to talk to another Devin and make a decision. There is no way that these things aren't soon going to be fully-functional employees. It's been fun to push the limits of running fully agentic software teams with @DaleSpoonemore who joined the @HelloUntangle team today. Our Devins are passing messages back and forth to us into Slack when needed and sometimes just talking session to session without us.

@thsottiauxOct 5

Day 1/ We have optimized the default speed to be ~50% faster across GPT-6 Astra and GPT-6.1 Sol through the subscription across all our products and partners using Sign in With ChatGPT (including OpenCode, Pi, Amp, Devin, ...). No changes needed on your end and this should be felt within the next two hours.

@theoOct 2

Things included in this PR: - Pi support. - Auto-resume when limits reset - T3 Code MCP (create, launch, message, wait on, read, search and interrupt threads) - `delegate_task` lets an agent start child agents on any provider or model - ACP Registry: add any registry agent (Devin, Cline, Kimi, Droid, and more) - Cursor runs on the official Cursor SDK, not the (broken) CLI - OpenCode 2 support - Agents can rename threads, regenerate titles, link PRs and settle threads - New MCP tools for worktree handoff, and for listing and closing previews - Proper server-side queueing and steering - Thread forking - Switch provider or model in the middle of a thread - Native subagents show as child threads in a lineage view, with their model, status and history - Use `@` with a thread name, or drag a thread from the sidebar, to attach it as context - Scheduled tasks - Tons more I'm forgetting Pumped this landed. We'll have a lot of cleanup to do to get it perfect, so expect lots of updates throughout the next few days :)

@thsottiauxSep 29

You can now use your ChatGPT subscription directly in over 16 partners products. No little rules, you can just use all your included usage right there. This includes things such as Devin, OpenCode, Notion, and many more. Sign in with ChatGPT and go.

@thsottiauxSep 29

You can now use your ChatGPT subscription directly in over 60 partners products. No little rules, you can just use all your included usage right there. This includes things such as Devin, OpenCode, Lovable, and many more. Sign in with ChatGPT and go.