Founder and CEO @ SolZero. Staff DevOps @ Consensys. Come hang out https://teapawt.org/t.co/q6mPm3fGl3
Canton, Georgia
Joined May 2023
- Tweets2.9K
- Following586
- Followers853
- Likes4.8K
Pinned Tweet
TLDR;
⚡ Fast, efficient custom harness built on @CloudflareDev
📦 Codex/OpenCode/CC sandboxes
🔁 Durable deterministic + agentic Workflows; session reuse/cache
🔐 Better Auth: OIDC/social/password
🧰 MCP, Skills, LiteLLM, Context Forge
⚙️ Alchemy IaC, EffectTS, Nub, Kumo UI
This is a major win for those trying to speed up CI. Only 2 of my 45 tests will run for this PR.
Also, s/o to @Cloudflare Clef decision model + @alchemy_run for local stack emulation to make for a delightful e2e test suite experience
LFGGGG!! 🔥 🔥 🔥
Time to build a better npm.
Just applied to @cloudflare for startups, courtesy of @atlas. I love building on Cloudflare so this is going to be very welcomed.
Bought SolKube.com earlier today, will show off the app soon.
Hey @michellechen, for the Models doc page can you add @ArtificialAnlys benchmark scores so I can compare performance? Also a cost/performance metric?
My favorite thing about Grok Bot is that it has solved coding from my phone. I’ve been chasing it for so many years, and it feels great
Been wanting to build a replacement for Lens (k8 management app) for a while and finally landed on something that I feel good about. Will be ironing things out over the next few days before I share more.
Here's some icon design exploration as a tease in the meantime.
'likelihood that you can and should merge' is an incredible way of comparing models, especially when considering cost.
E.g. Opus 5.5 is 59% more likely than Grok 4.7 to create a PR that is worthy to merge, and is slightly cheaper to do so.
Introducing FrontierCode 2.0: our new coding benchmark that measures a model’s ability to go beyond the spec.
Most coding evals measure a model’s ability to write code that meets a spec. With FrontierCode 1.0, we asked a new question: would you actually merge the PR?
FrontierCode 2.0 is comprised of more diverse tasks, featuring larger scopes and higher mergeability standards. We are now asking: not only would you merge this PR, but should you?
One key feature of FrontierCode 2.0 is underspecified prompts. Rather than including every engineering decision, prompts are deliberately ambiguous, giving the models room to demonstrate their judgment.
The result is a benchmark that is far from saturated, with the top model scoring just 45%.
Just applied to @cloudflare for startups, courtesy of @atlas. I love building on Cloudflare so this is going to be very welcomed.
Never been a believer in cultural appropriation, but seeing upper class kids rocking mullets may sway me. We used to put double wide kids in their place for trying that shit 😅
So clean and beautiful. From the grainy, semi-translucent border to the Liquid Glass details, its a pleasure to use.
Especially the git graph, a great way to what my agents are up to.
Introducing 1git
A native Git client for macOS. GPU-rendered (GPUIX-GPUI), Open-Source, and calm by design
Tabs for every repo. Local or SSH in one workspace. Multiple accounts support
here → git.1bye.dev
Here’s my Grok Bot for deep research.
For standard queries it uses Grok web search.
For complex tasks it uses @ExaAILabs, @CeramicSearch, and @Linkup_platform for rich insights that are token efficient. Requires a Cloudflare web search API key.
x.ai/bot/PkwpnZEg7KrFkd1PCPJ…
I created this Cloudflare admin bot to help
- review my spend, security, and perf weekly
- keep me and my bots up to date on Cloudflare news (its cool how proactive my bots are)
- clean up DNS, resources, tokens, etc
Here’s a template to set up your own
x.ai/bot/R8z7Gt048fEu5Z8xqtg…
Jonathan Beckman retweeted
Introducing 1git
A native Git client for macOS. GPU-rendered (GPUIX-GPUI), Open-Source, and calm by design
Tabs for every repo. Local or SSH in one workspace. Multiple accounts support
here → git.1bye.dev
Jonathan Beckman retweeted
Got Rust hot reload, even with GPUI. Without writing any additional code into app itself.
But currently initial build is slow, it took 195.62s to build this app from scratch to have hot-reload, with Zed it takes 800s+
Have to fix that. Btw it works kinda like Vite, by exposing it's own SDK to write integrations, that allows GPUI to survive state between hot-reloads
It's name is Rite (like Vite but instead of V is R)
Jonathan Beckman retweeted
You'd probably love this (when I met you last year you were expecting)...
I have a grok bot for each of my boys and I have it manage their knowledge base in notion. Recently we have been tracking the new words my son Thomas says and now, armed with that data, Thomas' grok bot generates new worksheets each week. Catered to his knowledge level.
So cool. This is the first iteration of it btw, can't wait to see what this turns into. I can probably send over a bot template if you're interested.
