Pinned Tweet
1/6 - You change the prompt. The answers look better.
Someone asks: “How do you know?”
“It feels better.”
That’s how a lot of LLM testing still works.
I’ve been exploring a more systematic approach 🧵👇
wait …. what?
New on our Engineering blog: Creating in-house source control.
In May, we merged around 2,500 code changes per day. In September, we merged more than 30,000 per day. As a result, we became bottlenecked by developer infrastructure built for human (rather than agent) scale.
Here, we share how we built a new version control system—and fully migrated our monorepo from git—in five months: anthropic.com/engineering/re…
46
Maybe Apple’s Vision Pro is starting to make sense
GPT-6 Astra + Unreal Engine 5.8 = real estate UX boost
I transformed 20 photos of a real property listing on OLX into a 3D interactive space you can walk inside.
For exterior, we used Google Maps panoramas with relevant coordinates
This elevates the customers' decision making process & boosts the sales funnel (without spending a thousands on just a walkable render)
Let me know if you want me to share the full process
This is HUGE!!! Claude Code is becoming more than just a coding assistant
Projects now run from one conversation, starting in Claude Code. You describe what needs doing, and Claude directs parallel threads that keep working after you close your laptop.
In beta today for select Pro and Max users in cloud sessions; coming to all Claude users soon.
1
82
What actually matters when evaluating an open model for self-hosting? Not the version number, but what changed inside and what that means for compute, memory, and hardware.
ugolattanzi.com/en/blog/enco…
Special thanks to @ivanfioravanti for the invaluable help in shaping and writing this piece 🙏
1
3
267
I haven’t written .NET code in a while, but the Performance Improvements post by @stephentoub that drops with every release is still an absolute MUST-read.
Performance Improvements in .NET 11
devblogs.microsoft.com/dotne…
1
90
5/6 -And where do new test cases come from?
Production.
Your traces already contain real scenarios, edge cases, and failures.
Turn them into tests.
👉 ugolattanzi.com/en/blog/from…
1
37
6/6 -The goal isn’t to test everything with an LLM.
It’s to use the cheapest reliable test at each layer.
I put the implementation here:
👉 github.com/imperugo/how-to-t…
How are you testing your LLM applications today?
1
40
Highly recommended, especially if you work in an Enterprise environment
Deep Dive - Caching in .NET is finally here 🥳
After a full year of hard work, my 2nd course is out on @dometrain (30% off with "AUGUST30").
It's the conclusion to the journey we started with the 1st course, Getting Started.
I condensed 20+ years...
dometrain.com/course/deep-di…
Am I the only one who feels like I'm losing something with the continuous release cycle of @ClaudeDevs ? I'm a fan of the approach, but there's a thin line where it moves from "good" to "hard to keep up with".
1
45
Context compaction is not the fix, it is the symptom
ugolattanzi.com/en/blog/cont…
1
40
/handoff by @mattpocockuk is the perfect solution, but you have to know when to run it
23









