AI is quietly eating two very different worlds this week: software development and pure mathematics.
Cursor's side chats and agent history are small UX fixes, but they signal something bigger: agents are becoming persistent collaborators, not just one-shot tools. Meanwhile, GPT-5.6 Sol Ultra proved a 50-year-old math conjecture using 64 parallel subagents in under one hour. Same multi-agent pattern, wildly different stakes.
The throughline? Parallelism is the new intelligence.
|
|
In Partnership with
|
·
|
|
|
|
|
|
Today's Author
Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.
|
|
Top Repo
|
|
|
|
OpenAI's GPT-5.6 Sol cracks a 50-year-old math conjecture in under one hour
|
|
6,634 Likes
|
|
GPT-5.6 Sol Ultra just did something no human mathematician could pull off in 50 years. It produced a proof of the Cycle Double Cover Conjecture, a famous unsolved math problem, in under one hour. So what even is this conjecture? Think of any network of nodes and connections with no single critical link. The conjecture asks: can you always find a set of loops that covers every connection exactly twice? Mathematicians said "probably yes" since the 1970s. Nobody could prove it. Until now. Here's how it worked: - The model ran 64 subagents in parallel, meaning 64 copies working simultaneously on different angles of the problem
- OpenAI published the full prompt, the proof PDF, and a Lean formalization (a machine-checkable version of the proof) on GitHub at openai/cdc-lean
- The proof is not yet peer-reviewed, so the math community is still verifying it
What this opens up: you can point this model at hard, open-ended research problems using a multi-agent setup. The prompt is public, so you can study exactly how to structure that kind of task yourself.
|
|
|
|
Presented by Granola
|
|
This App Might Actually Make You Love Meetings
|
|
Meetings aren't the problem. It's everything that comes after. Granola, the AI Notepad for people in back-to-back meetings, lets you avoid the stress of things slipping through. Take notes like you always have. Granola works in the background, turning conversations into clear summaries, action items, and next steps. Before, during, or after, chat directly with your notes to catch what you missed and know what happens next. Try Granola free for a month with code ALPHASIGNAL.
|
|
|
|
Top News
|
|
|
|
Cursor ships searchable agent history and side chats for parallel conversations
|
|
4,974 Likes
|
|
Cursor just shipped v3.11, and it fixes one of the most annoying things about working with AI agents: losing track of your conversations. Here's what's new: - Agent transcript search - Cursor builds a local index of all your past chats. You can now search across thousands of conversations instantly. Inside any open chat, hit Cmd+F to jump between matches with a match counter.
- Side chats - Open a parallel conversation without killing your main thread. Use /side or /btw to spin one up. Each side chat is a full agent conversation you can revisit later, and you can @-mention it to pull its context back into your main chat.
- Cloud agent hooks - You can now hook into the agent conversation itself: prompts, responses, thinking, and turn completion. This lets you build self-correcting agent loops.
- Smarter project picker - Connect GitHub, GitLab, or Azure DevOps without ever leaving the picker. Fewer clicks to launch agents.
Side chats are the real unlock here. Explore a tangent, investigate a bug, then bring the answer back. No more cluttered main threads.
|
|
|
|
Presented by Bright Data
|
|
The web trained LLMs. Now it's training robots.
|
VLAs (Vision-Language-Action models) need web-scale video data to handle new contexts. At the MACHINA Summit last week, Bright Data's CEO showed how thousands petabytes of open-web visual data train humanoid robots - now searchable by prompt and delivered pipeline-ready.
|
|
|
|
Top News
|
|
|
|
Google AI Studio lets deployed apps claim a permanent custom domain
|
|
3,099 Likes
|
|
You built an app in Google AI Studio. You hit deploy. And then you get a URL like burning-man-animal-cuddle-614365371127.us-west1.run.app. Not exactly portfolio material. Google just fixed that. AI Studio now lets you pick a custom subdomain under ai.studio, so your app lives at something like your-app-name.ai.studio instead of a random word salad. Here is what you need to know: - Set it inside the publishing menu when you deploy, or update it later
- Names are unique across all of AI Studio, first-come first-served, so grab yours fast
- If you unpublish your app, the URL gets released and anyone can claim it
- You can change it later, but the old address becomes available to others immediately
This makes sharing demos, building a portfolio, or sending a link to a client actually feel professional. Your app finally has a real identity.
|
|
|
At Alpha Signal, our mission is to build a sharp, engaged community focused on AI, machine learning, and cutting-edge language models, helping over 200,000 developers stay informed and ahead. We're passionate about curating the best in AI, from top research and trending technical blogs to expert insights and tailored job opportunities. We keep you connected to the breakthroughs and discussions that matter, so you can stay in the loop without endless searching. We also work closely with partners who value the future of AI, including employers and advertisers who want to reach an audience as passionate about AI as we are.
Our partnerships are based on shared values of ethics, responsibility, and a commitment to building a better world through technology.Privacy is a priority at Alpha Signal. Our Privacy Policy clearly explains how we collect, store, and use your personal and non-personal information. By using our website, you accept these terms, which you can review on our website. This policy applies across all Alpha Signal pages, outlining your rights and how to contact us if you want to adjust the use of your information. We're based in the United States. By using our site, you agree to be governed by U.S. laws.
|
|
|
Looking to promote your company, product, service, or event to 250,000+ AI developers?
|
|
|
|
|