Launch HN: Manufact (YC S25) – MCP Cloud
Stitch together MCP server, React UI, hosting, auth, and scaling. npx create-mcp-use-app scaffolds the whole stack. Test by connecting to a live LLM and hoping. Visual Inspector. Sandbox testing. No LLM required. Manage your own deployment pipeline and branch previews. git push auto-deploys via the GitHub App. Build for ChatGPT or Claude or Gemini — […]
Show HN: Mail Memories – A desktop app to rescue photos from Gmail
The Problem Existence isn’t ownership. Technically, your photos are in your inbox. But practically? They’re gone. A cherished memory buried under decades of newsletters, receipts, and spam isn’t something you have—it’s just data. If you can’t find it, view it, or share it, it might as well not exist. Mail Memories was built to bridge […]
Show HN: ZeroFS – A log-structured filesystem for S3
$ nbd-client 10.0.1.5 10809 /dev/nbd0 -N storage -persist # us-east $ nbd-client 10.0.2.5 10809 /dev/nbd1 -N storage -persist # eu-west $ nbd-client 10.0.3.5 10809 /dev/nbd2 -N storage -persist # ap-southeast $ zpool create global-pool mirror /dev/nbd0 /dev/nbd1 /dev/nbd2 $ zpool status global-pool | grep state state: ONLINE
How to ask for help from people who don’t know you
30 June 2026 No matter what you’re doing, from building a civilization on Mars to getting a summer internship, you will have to ask people for help. Yet, most people get this crucial skill wrong. They put themselves at the front of their request, when they should be putting the other person there. But isn’t […]
Is One Layer Enough? A Single Transformer Layer Matches Full-Parameter RL Train
Abstract:Reinforcement learning (RL) has become a central component of post-training large language models (LLMs), yet little is understood about how RL adaptation is distributed across transformer layers. Existing approaches typically update all model parameters uniformly, implicitly assuming that every layer contributes similarly to the gains obtained during RL post-training. In this work, we challenge this […]
Vite+ Beta
TL;DR: Vite+ is now in beta. It unifies the runtime, package manager, and frontend tools every web project needs behind one fast, consistent workflow. Start a new project with vp create, or adopt it in an existing project with vp migrate. Today, we are releasing the beta of Vite+, the unified toolchain for the web. […]
PeerTube is a free, decentralized and federated video platform
Website | Join an instance | Create an instance | Chat with us | Donate Be part of a network of multiple small federated, interoperable video hosting providers. Follow video creators and create videos. No vendor lock-in. All on a platform that is community-owned and ad-free. Developed with ❤ by Framasoft Introduction PeerTube is a […]
The fall of the theorem economy
How AI could destroy mathematics and barely touch it “The product of mathematics is clarity and understanding. Not theorems, by themselves.”—Bill Thurston Handwritten diagram by Alexander Grothendieck My best theorem is one I never wrote down. It crystallized one bright morning in Lausanne, Switzerland, as I was preparing for my last invited conference talk. The […]
CursorBench 3.1
We evaluate agents on ambiguous, multi-file tasks from real Cursor sessions. Higher scores are better. More about CursorBench↗ A scatter and line chart comparing Fable 5, Opus 4.8, Opus 4.7, GPT-5.5, Sonnet 5, Sonnet 4.6, GLM 5.2, Composer 2.5, and Composer 2 scores against average cost per task.75% CursorBench 3.1 score70%65%60%55%50%45%$20$16$12$8$4$0Average cost per taskFable 5 […]
Kimi K2.7 Code is generally available in GitHub Copilot
Release July 1, 2026 • 1 minute read Table of Contents Menu. Currently selected: Availability in GitHub Copilot Kimi K2.7 Code, an open-weight model, is now generally available in GitHub Copilot. This is the first open-weight model offered as a selectable option in the Copilot model picker, giving you more choice and a lower-cost option […]