Tom MacWright
@macwright.com
4 days ago big news as soon as i fix and merge this revert revert revert revert branch
💬 2
♻️ 1
❤️ 68
big news as soon as i fix and merge this revert revert revert revert branch
I spent the weekend riding California's Lost Coast! Then wrote a lil blog post about it :)
ellie.wtf/posts/the-lo... atuin v18.20 is out now! 🐢
- sync is 32x faster and uses way less bandwidth
- the installer now logs you in, so a new machine (or sandbox) has your full history almost instantly
- syntax highlighting on Windows, PowerShell, and Illumos
read all about it: blog.atuin.sh/making-atuin... Our launch post walks through registering your first WebMCP tool and links all five demo apps.
www.netlify.com/blog/compete... The full rules, judging criteria, and submission instructions are on OpenAI's page. Sean Roberts from our team is on the judging panel.
openai.com/webmcp-challenge We published five demo apps you can read and fork, including a marketplace with a working checkout flow and a restaurant where your agent hits a genuinely full time slot and has to negotiate.
The form is open to everyone in the challenge. The first 1,000 builders to fill it out each get 3,000 Netlify credits, which cover building and running the app. Every project that lands in the top 10 gets an additional $500 from Netlify.
WebMCP is an experimental open web standard that lets a website expose structured tools an AI agent can call directly, instead of making it click and type through an interface built for a person with a mouse. ChatGPT's in-app browser supports it out of the box.
Netlify is partnering with OpenAI to support the WebMCP Challenge, which opens today. Build an agent-native web app, or add WebMCP to one you already run!
There are 3 million Netlify credits up for grabs and a $5,000 prize pool for builders:
webmcpchallenge.netlify.app The recording of our live session, “Using Elicit to Make More Rigorous Decisions”, is now available www.youtube.com/watch?v=KBS...
In the session, Kadir Annamalai, product lead for the Elicit Research Agent, walks through a live, multi-step workflow showing how Elicit can support decision-making. The Next.js team disclosed two critical vulnerabilities today. Both can lead to unauthenticated remote code execution. Netlify sites are not affected, but we still recommend upgrading to 15.5.24 or 16.3.3.
www.netlify.com/changelog/20... A large table is never just a large table. It's bloated. Slow to vacuum and repack. Holds connections. Over-indexed. It's a big pain that sharding solves.
the protocol labs narrative is starting to solidify into it having been a very spendy labs project that ultimately benefited atproto and iroh instead of its own projects ipshipyard.com/blog/2026-th... Paste images as your chat input.
• Take a screenshot
• Paste it into the chat
• Describe what’s wrong
• Watch Command Code fix the issue
Let your screenshots be the context. Netlify Agent Runners keeps picking up models, including a growing collection of open models. Whichever one you pick, it can ask you clarifying questions when useful, and it delivers its work as a deployed application.
www.netlify.com/blog/new-cla... Bay Area Rust’s next meetup is all about graphics, with talks, food, and time to meet other Rust developers.
Join us at @atuin.sh's SF office on September 16, 6:00–8:30 PM.
Interested in giving a talk? Give Esther a shout through the event page.
luma.com/9oiujuyw But seriously, we’re so grateful to you and everyone else showing up and giving our repo some love. Only a few more weeks left in the Zed Summer Repro Program. 🏃🏻
Benjamin van Pham and Aidan Van Duyne each confirmed one: a third-party extension crash and an unusual gutter-width bug. Paul won the weekly raffle for the third time! Slow down, Paul! 😂
Paul Contreras and Carlos Urias confirmed 6 bugs each. Paul caught inlay hints silently failing on hover. Carlos dug up a Git gutter scaling bug.
@lamersc.com found an easy-to-miss interaction between Prettier and hard tabs. Zero confirmed bugs, but his five non-reproductions were still some of the most useful work on the list. Learning that a report does not reproduce in another environment gives our developers more context and helps narrow down the conditions that trigger it.
6 Student Ambassadors, 22 reproduction attempts, 11 reproduced, 11 not reproduced. Daniel Paredes ran 5 reproduction attempts during week four, and every one came back “could not reproduce,” including a Windows file-deletion dialog bug that simply wasn’t occurring in his setup.
good history of monsignor mcgolrick park in brooklyn, fun to presume that if he had kids they would've hated naming that park after him