Moritz Stephan (@at_code_wizard) 's Twitter Profile
Moritz Stephan

@at_code_wizard

@cognition_labs | @StanfordAILab

ID: 1387778698863054848

calendar_today29-04-2021 14:40:06

54 Tweet

890 Followers

147 Following

Adhyyan Sekhsaria (@adhyyan1252) 's Twitter Profile Photo

Kinda insane to work on QA Devin. Some of the code for this was written by Devin. It was Devin writing code to get Devin to start other Devins!

Bruno F | Magna (@bfaviero) 's Twitter Profile Photo

24h with Devin and ok our team actually LOVES it. Already been asked to increase our budget for it 😂 Useful for small things w/ little supervision, but also kicking off big things– Scaffolding code, finding relevant files/functions, and an engineer finishes it off. And we're

Moritz Stephan (@at_code_wizard) 's Twitter Profile Photo

Definitely on the silly side – but I also gave it some examples of prior gifts & these fit pretty well (socks with famous art pieces etc.)

Definitely on the silly side – but I also gave it some examples of prior gifts & these fit pretty well (socks with famous art pieces etc.)
Cognition (@cognition_labs) 's Twitter Profile Photo

Did you know that the average PR waits 100+ hours to get reviewed? Watch how to use Devin’s API to automate PR reviews, cutting days to minutes. Here’s how to set up Devin-powered PR reviews on your repos in <5 mins 🧵👇

Mercor (@mercor_ai) 's Twitter Profile Photo

mercor, Etched, and Cognition are hosting a 24-hour inference-time compute hackathon on 2/28 with $60k in cash prizes. apply at mercor.com/hackathon

mercor, <a href="/Etched/">Etched</a>, and <a href="/cognition_labs/">Cognition</a> are hosting a 24-hour inference-time compute hackathon on 2/28 with $60k in cash prizes. apply at mercor.com/hackathon
Cognition (@cognition_labs) 's Twitter Profile Photo

📢 Join us for the Inference Time Compute Hackathon, hosted with @etched and @mercor, sponsored by @Coreweave and Paul Jankura! Teams will push the limits of reasoning and compete for the grand prize of $40k cash + more. Apply at mercor.com/hackathon

📢 Join us for the Inference Time Compute Hackathon, hosted with @etched and @mercor, sponsored by @Coreweave and <a href="/Anthropic/">Paul Jankura</a>!

Teams will push the limits of reasoning and compete for the grand prize of $40k cash + more.

Apply at mercor.com/hackathon
Moritz Stephan (@at_code_wizard) 's Twitter Profile Photo

Just shipped Devin project templates 🚢! These are pre-configured repos for building apps from scratch with Devin. They're designed to be simple to run & test locally, so Devin has a built-in feedback loop. If you ask Devin nicely, it will even deploy your apps Might add more

Just shipped Devin project templates 🚢! These are pre-configured repos for building apps from scratch with Devin. They're designed to be simple to run &amp; test locally, so Devin has a built-in feedback loop. If you ask Devin nicely, it will even deploy your apps

Might add more
Moritz Stephan (@at_code_wizard) 's Twitter Profile Photo

deep integration into every part of the engineering workflow is the future of swe agents. Excited to start building that out with our new Linear integration (more to follow soon). My workflow: - Plan out a large project with Devin Search & create tickets - Use the Devin linear

Ryan Bai (@ryanbai1412) 's Twitter Profile Photo

Just shipped a new knowledge page in Devin. You can now sort your knowledge into folders and perform bulk actions. You can also have Devin organize your knowledge for you!

Just shipped a new knowledge page in Devin. You can now sort your knowledge into folders and perform bulk actions. You can also have Devin organize your knowledge for you!
Sunflower Capital (@seedtosunflower) 's Twitter Profile Photo

We’re excited to announce Sunflower Capital Funds I and II. Sunflower is a $250m fund that partners at the earliest stage with companies building foundations for modern enterprises, critical industries, and the physical world.

We’re excited to announce Sunflower Capital Funds I and II. Sunflower is a $250m fund that partners at the earliest stage with companies building foundations for modern enterprises, critical industries, and the physical world.
Cognition (@cognition_labs) 's Twitter Profile Photo

Our research interns present: Kevin-32B = K(ernel D)evin It's the first open model trained using RL for writing CUDA kernels. We implemented multi-turn RL using GRPO (based on QwQ-32B) on the KernelBench dataset. It outperforms top reasoning models (o3 & o4-mini)! 🧵

Our research interns present:
Kevin-32B = K(ernel D)evin

It's the first open model trained using RL for writing CUDA kernels. We implemented multi-turn RL using GRPO (based on QwQ-32B) on the KernelBench dataset.

It outperforms top reasoning models (o3 &amp; o4-mini)! 🧵
Moritz Stephan (@at_code_wizard) 's Twitter Profile Photo

I've spent countless hours looking at the traces of my Devins. Seeing when the agent fails and why helped me write better prompts and understand the right task size it can complete reliably. Excited to make more of our internal tools available to all users.