Copilot Platform @GitHub. Formerly Observability @Honeycombio and F# lang/.NET tools @Microsoft
Redmond, WA
- Tweets 7,650
- Following 359
- Followers 5,713
- Likes 22,507
Phillip Carter retweeted
There is no silver bullet or perfect prompt but if you do hard core engineering (not related to LLMs themselves but instead how you apply them) the craft gets recursive very very fast.
People keep asking me about what skills I use and how I prompt LLMs and the answer shocks folks.
I don’t pre-install any skills or rely on attempting to eval max markdown into a context window. There’s no magic trick except application of the craft of software engineering itself!
Stop chasing LLM prompt fairies and instead focus on the craft.
21
18
195
9,940
The typescript calls the rust which calls the go which calls the go which calls the rust then it all goes back
Tada, you have a cloud platform for agents!
1
4
527
Am I missing something about Muse? Everything just has like half access or makes me plug in API tokens
1
696
Phillip Carter retweeted
Messing around with @typesafeai Jev for real-time styling
🔤 🔜 🎨
If you can define a bounded, semantically meaningful property space, you can project a natural language expression into that space in one generation step
3
14
218
12,564
Extremely detailed post on migrating a sophisticated codebase to ~800k lines of Rust through collaborative agents and a different “human in the loop” approach m. Worth a read with your morning coffee! github.blog/ai-and-ml/genera…
6
2
29
2,299
Treat AI like going to the bathroom in the office where you mosey on in and end up relieving yourself next to your boss’s boss’s boss who is also relieving themselves and you both hear each others’ noises but do not acknowledge each others’ presences
We need to treat using AI like going to the bathroom: I know you do it, but you shouldn't talk about it in polite company, and I should never see the output.
1
408
So anyways, I'm pro-"Pacing the Frontier" not because of some existential scenario, but simply because it's untenable to push models whose actions we can't even trace anymore (Astra), let alone understand.
3
3
774
The whole x-risk thing just doesn't spook me as much, even knowing that cybersecurity systems are in existential danger. It's just that all the other scenarios involve massive human and physical world constraints that aren't bottlenecked on compute.
1
2
160
Real-SWE might be the first coding benchmark I’ve seen in a while where the scores are actually close to my lived experience with SOTA models withspecific.com/benchmarks/…
3
769
So it’s a new one each day huh rubyhack.ai/
4
703
We’re doing some great work around PRs, try the new stuff out now and be prepared for more goodness soon :)
A much-needed refresh of the repo pull requests page landed in public preview yesterday. Very much in line with the rework we did on github.com/pulls earlier this year: queries that can do AND/OR and nesting, content assist while you type, a filter sidebar, a compact mode, and check status and unread updates inline. We've all got a deluge of PRs these days, so we're working to make them easier to manage.
github.blog/changelog/2026-0…
Try it out and give us feedback!
1
471
