The dev tooling layer is quietly being rebuilt around agents, not humans. Warp open-sourced its terminal with an agent-first contribution model. Cursor just broke its agents out of the editor entirely with a new SDK.
The pattern is clear: agents are becoming infrastructure, not features. Your terminal, your CI pipeline, your IDE , all becoming agent runtimes.
And if the Abstract Chain-of-Thought paper lands, those agents will think 11x cheaper. Good week to be building.
|
|
Together with
|
·
|
|
|
|
|
|
Today's Author
Lior Alexander. Founder of AlphaSignal and former ML Engineer. Previously built ML systems at Iguazio, Guesty, Enphase, Mila.
|
|
Top News
|
|
|
|
Warp terminal goes open-source with GPT-powered agents handling contributions
|
|
8,672 Likes
|
|
Warp, the AI-powered terminal, just went open source. The full Rust codebase is now on GitHub, and it already crossed 37,000 stars within hours of launch. Here is what actually changed and what you can do with it now: - The code is yours to read, fork, and modify under an AGPL v3 license, meaning any changes you ship publicly must also stay open source.
- The UI framework is under the more permissive MIT license, so you can reuse it freely in your own projects.
- You can plug in your own AI agent: Claude Code, Codex, Gemini CLI, or Warp's built-in one all work inside it.
- Contributions work differently here. AI agents handle the actual coding. You focus on proposing ideas, writing specs, and reviewing results.
- Use the new /feedback skill inside Warp to open GitHub issues without leaving your terminal.
To contribute, read CONTRIBUTING.md in the repo. OpenAI is the founding sponsor, and GPT models power the agent workflows behind the scenes.
|
|
|
|
Presented by Viktor
|
|
An AI coworker with access to your whole stack.
|
|
Yesterday, engineering asked Viktor to review open PRs on GitHub, cross-reference Linear tickets, and flag anything blocking the sprint. Posted to #dev before standup.
Finance asked Viktor to pull Stripe revenue and compare it to Google Ads spend. Formatted PDF in #finance by morning.
Same Slack workspace. One colleague across 3,000+ integrations with read/write access. Persistent memory. No workflows to build.
9,000+ teams. SOC 2 certified. Your data never trains models.
|
|
|
|
Top News
|
|
|
|
Cursor releases open-source SDK to build and deploy coding agents anywhere
|
|
7,732 Likes
|
|
Cursor just shipped the Cursor SDK, and it changes what you can actually build with AI agents. Until now, Cursor's agents only lived inside the Cursor editor. They were locked to the editor. Now they can run anywhere you need them. So what does that unlock? - Set up an agent to pick up a bug report, write a fix, and open a pull request, all without anyone stepping in.
- Agents can run automatically when tests fail in a CI pipeline, summarize code changes, or help non-technical teams query data.
- Switch between any model Cursor supports, optimizing for cost and capability with a simple config change.
- Run agents locally or on Cursor's cloud, where each gets its own virtual machine with the repo already cloned.
Getting started is simple. Install via npm install @cursor/sdk, and billing is token-based. Cursor also released starter projects on GitHub, including a kanban board and a CLI tool for spinning up agents from a terminal. Companies like Rippling, Notion, Faire, and C3 AI are already using it. The SDK is in public beta now.
|
|
|
|
Presented by Brave
|
|
The search API your agent actually needs
|
|
40B pages. Independent index. $5/1K calls. Brave Search API gives your chatbots and agents real-time web data - no Google scraping, no middlemen.
|
|
|
|
Top Paper
|
|
|
|
New training method cuts LLM reasoning tokens by 11x using abstract token chains
|
|
1,003 Likes
|
|
AI models think out loud. That's expensive. Every reasoning step gets written in plain English, token by token, before you get an answer. More tokens = more cost = slower responses. Abstract Chain-of-Thought flips this. Instead of thinking in English, the model reasons through a short sequence of special reserved tokens, like a private shorthand language it invents for itself. You still get a normal English answer at the end. The results are pretty striking: - 11.6x fewer reasoning tokens used during inference
- Matches or beats standard verbal reasoning on math, multi-hop questions, and instruction-following
- Works across different model families, not just one specific model
What makes it work is a two-stage training process: first the model learns what these abstract tokens mean by peeking at real reasoning chains, then reinforcement learning sharpens them further. The big unlock here is cheaper inference without sacrificing quality. Faster responses, lower API costs, same accuracy.
|
|
|
|
Signals
|
|
|
|
|
| 2 |
|
You think 4x faster than you type. Wispr Flow closes the gap inside Cursor, Claude, and ChatGPT. 89% zero edits.
|
| Presented by Wispr Flow |
|
|
|
|
|
|
|
|
|
|
|
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?
|
|
|
|
|