David Thyresson

Ellie Huxtable @ellie.wtf

115 PRs to Zed. A 60+ component library. A production app that replaced Electron. All centered around GPUI, Zed's GPU-powered UI framework.

bsky.app/profile/huac... is a Zed Community Champion 🏆

zed.dev/blog/community-champion-jason-lee

💬 2 ♻️ 1 ❤️ 43

CommandCode.ai @commandcode.ai

feeling this so hard rn

💬 1 ♻️ 0 ❤️ 4

AI Terminology #28: Beam Search

↳ A decoding algorithm that explores multiple possible output sequences in parallel and keeps the most likely ones at each step to improve overall results.

💬 0 ♻️ 0 ❤️ 0

one of those social networks that creates ghost profiles for people but also makes ai slop clown avatars as incentive to sign up and replace it

💬 1 ♻️ 0 ❤️ 15

Netlify @netlify.com

GPT-5.5 and GPT-5.5 Pro are live in Netlify AI Gateway and Agent Runners.

Zero config. No API keys to manage. Just import the OpenAI SDK and ship.

www.netlify.com/changelog/gp...

💬 0 ♻️ 0 ❤️ 0

an app that uses atproto blobs as 'generic file storage' seems eminently possible but maybe against the intent of the system?

💬 4 ♻️ 0 ❤️ 15

Get all the details in our latest blog 👇
pris.ly/ts-migratio...

💬 0 ♻️ 0 ❤️ 0

A migration fails halfway through. What already ran?

In Prisma Next, migrations are TypeScript files with checks before and after every operation. 

You can safely edit, re-run, and fix the exact step that failed.

💬 2 ♻️ 0 ❤️ 2

Check it out!
pris.ly/ppg-backup

💬 0 ♻️ 0 ❤️ 0

Need to create a manual backup of your Prisma Postgres database?

pg_dump works great with direct connections:

💬 1 ♻️ 0 ❤️ 0

been eyeing tunaformac.com as an alternative, i already use and absolutely love the creator's other launcher, leaderkey

💬 1 ♻️ 0 ❤️ 4

i've passively appreciated @redblobgames.com for years but this is the first time i've dedicated the time to focusing and reading his writing. it's an absolute treasure, an incredible resource on so much, and incredibly high quality. for example an intro to a*: www.redblobgames.com/pathfinding/...

💬 2 ♻️ 1 ❤️ 23

confirmed that i still got that dog in me

💬 4 ♻️ 0 ❤️ 25

observable podcast! www.youtube.com/watch?v=AUh5...

💬 0 ♻️ 0 ❤️ 10

worst case

why it cannot be worse

why it cannot be better

$Worst case In the worst case, A* expands all nodes n {\textstyle n} for which f ( n ) = g ( n ) + h ( n ) ≤ C ∗ {\textstyle f(n)=g(n)+h(n)\leq C^{*}}, where C ∗ {\textstyle C^{*}} is the cost of the optimal goal node. Why it cannot be worse Suppose there is a node N ′ {\textstyle N'} in the open list with f ( N ′ ) > C ∗ {\textstyle f(N')>C^{*}}, and it's the next node to be expanded. Since the goal node has f ( g o a l ) = g ( g o a l ) + h ( g o a l ) = g ( g o a l ) = C ∗ {\textstyle f(goal)=g(goal)+h(goal)=g(goal)=C^{*}}, and f ( N ′ ) > C ∗ {\textstyle f(N')>C^{*}}, the goal node will have a lower f-value and will be expanded before N ′ {\textstyle N'}. Therefore, A* never expands nodes with f ( n ) > C ∗ {\textstyle f(n)>C^{*}}. Why it cannot be better Assume there exists an optimal algorithm that expands fewer nodes than C ∗ {\textstyle C^{*}} in the worst case using the same heuristic. That means there must be some node N ′ {\textstyle N'} such that f ( N ′ ) < C ∗ {\textstyle f(N')<C^{*}}, yet the algorithm chooses not to expand it. Now consider a modified graph where a new edge of cost ε {\textstyle \varepsilon } (with ε > 0 {\textstyle \varepsilon >0}) is added from N ′ {\textstyle N'} to the goal. If f ( N ′ ) + ε < C ∗ {\textstyle f(N')+\varepsilon <C^{*}}, then the new optimal path goes through N ′ {\textstyle N'}. However, since the algorithm still avoids expanding N ′ {\textstyle N'}, it will miss the new optimal path, violating its optimality. Therefore, no optimal algorithm including A* could expand fewer nodes than C ∗ {\textstyle C^{*}} in the worst case.$

💬 1 ♻️ 1 ❤️ 8

GPT-5.5 and GPT-5.5 Pro just landed in Zed. 💥 

Student plan: GPT-5.5 Pro plan: GPT-5.5, GPT-5.5 Pro

BYOK: both models available in v0.233.10 Stable / v0.234.6 Preview

zed.dev/releases

💬 2 ♻️ 1 ❤️ 26

Railway @railway.com

Red deploy? Let the Agent turn it green

Changelog #0287

• Railway Agent for every workspace
• Outbound IPv6 egress
• Fixes and improvements

railway.com/changelog/2026-04-24-railway-agent

💬 0 ♻️ 0 ❤️ 2

luma.com/9j3q5ejl

💬 0 ♻️ 0 ❤️ 2

Rust friends in the Bay, come hang with us on May 20th! 🦀 We're co-hosting the Bay Area Rust meetup at the Root VC office with talks from Servo, Atuin, and Zed.

Link in comments to RSVP.

💬 2 ♻️ 2 ❤️ 20

we talk about 'enshittification' a lot nowadays but i think the phenomenon for github is more akin to a corporate death-drive, the instinct of a company that hates beauty extinguishing the beautiful thing it acquired to spite the world

💬 1 ♻️ 2 ❤️ 40

HOWEVER

💬 0 ♻️ 0 ❤️ 1