I built https://t.co/R1jAMUfNTv — a free public database for American citizens who deserve safer communities.
You can track which judges released defendants who then got rearrested, skipped court, or violated their release conditions. All public records. All free.
I started with Orange County FL and will be expanding to all 67 Florida counties and eventually every state in the country. This first batch of info is from 2024 and since public reports are released in March/April for the previous year, data is behind. But I wanted to see if this is plausible. After adding 2024,I'll add 2025 and then figure out how to get real-time-data uploaded.
It's in beta — would love to know what you think 👇
Numbers don't lie, but criminals do.
https://t.co/DfTcJ6XMYn
@bennyjohnson@jockowillink@GrantCardone@LauraLoomer@nickshirleyy@j_fishback
Our teams have been building with a 2.5x-faster version of Claude Opus 4.6.
We’re now making it available as an early experiment via Claude Code and our API.
Introducing Claude Opus 4.5: the best model in the world for coding, agents, and computer use.
Opus 4.5 is a step forward in what AI systems can do, and a preview of larger changes to how work gets done.
My favorite time of year is back, and I’m giving away four annual TradingView subscriptions.
Eligibility:
1️⃣Follow @CarpeNoctom
2️⃣RT this tweet
Winners will be selected on Wed night EST (11/26) and notified via DM
Good luck!
#ad#blackfriday
So, I didn't renew my $100/month Claude subscription last month; I switched to the regular paid plan. Two days ago, I got a very annoying bug that no LLM could solve, and I was unlucky enough to max out my Claude usage, so it told me to fuck off until today at 10 pm.
I decided to be smart and wait for Gemini 3 Pro to be released, which was planned for today.
In the meantime, I tortured Gemini 2.5 Pro, Kimi, and ChatGPT for almost two days (in parallel with some other work) trying to make them fix the issue. No luck.
Today, after the disappointing performance of Gemini 3 Pro on this problem (it's as dumb as 2.5 Pro in code debugging as I already posted today), I finally returned to the $100/month Claude plan. The problem was solved after 15 minutes of interactive debugging.
The stubbornness of Google in rejecting to use man-made debugging sessions as training data is astonishing.
@Google , I love you and you are a big part of my digital life, but no one fucking cares how well you write English prose or generate working prototypes. Software engineering is where LLM's most lucrative (and beneficial for humanity) use case is, and you will fail it if you don't act as an adult.
Give your software engineers a choice to work on their own projects, no matter what they are, and make them record annotated debugging sessions, which include the LLM:
1. Including console logging lines into the buggy code and users (or tools) sending them back.
2. Injecting JavaScript code into the running app and the users (or tools) sending the browser console logs back to the LLM.
3. Creating a Python script that searches for something in the codebase and getting the search results back.
4. Doing many other application-specific debugging scenarios, including adding breakpoints and running unit tests.