Code editors are about to have their Xerox mouse moment. Engineers are stuck in the terminal era of computing. AI is catalysing a huge shift in what “coding” will look like. Here's what's about to happen

I’m hiring SWEs as we’ve got a huge surge in demand to deliver custom AI agents/data pipelines. People already joining include the Head of AI from a AI Unicorn, a super senior national AI advisor, MLEs from FAANG.

Everyone says VCs spread news fast but I found something faster. We started talking to one enterprise about using our synthetic reasoning traces, 3 of their competitors emailed within a week.

We eat our own dogfood Cosine. We're doing a whole week this week where we're not allowed to even open VsCode. This week only Genie builds Genie.

Amazing to work with the amazing GTM team at OpenAI

trygenie.now 🤫

No complicated compute units, no tabbing. It cost us less than $140 to make $88k on Upwork tasks. Special thanks to OpenAI/OpenAI Developers for working with us to turn an unique GPT 4.1 ver into one that is capable of reasoning, navigating and planning like a human dev, in any repo

When can we talk about the enterprise customer in banking that has the most legacy of legacy codebase?

Giving customers even more value per $ spent.

Why risk using models that blackmail you when you could be secure with Genie for a fraction of the cost and better performance?

Full-circle moment – I found the original doc with the idea for Genie back from July '22 (the davinci-2 days). Back then E2E projects were nearly impossible, by contrast now I can get Genie to 0-shot me a platformer and deploy it. Would have seemed like witchcraft back then.

Genie just got a major homepage upgrade: ➡️ Reopen recent tasks fast ➡️ Start with images + tickets ➡️ Add context before prompting ➡️ Pick your project upfront ➡️ Unsent drafts now auto-save ➡️ And more... Less friction. More flow. Discover here: go.cosine.sh/new-genie-home…

I think the Slack integration is underrated. Literally saw a colleague just tag Genie in chat after discussing what to build. And before the chat finished the PR was done

Building Genie outside of user's devices and their IDEs was a difficult decision. But seeing our own teams along side big banks, large media and healthcare companies all work fully asynchronously is still unbelievable to me. This isn't "background" as an experiment, it's fully

Working with OpenAI teams to customise and build with their o3 model has allowed us to turn a dream feature into reality. We're going to release all the details this Thursday, but for now, all I can say is that you may never need a PM or IDE ever again. The slash in costs means

Big shout out to our enterprise and on prem customers that helped test AutoPM at scale, in complex codebases and making us work hard to reach their high standards. Those Helm charts, VLLM configs and trying your entire Jira backlog in one go was worth it

We should redo live demos on stage with the help of Sarah Urbonas and shyamal again haha