Leo Boytsov

Europeans - here is how to avoiding tipping in the US without making waiters angry 1. Create a disposable virtual debit card (like Revolut) 2. When the bill comes, you usually pay for the meal first. Use that card to pay for it 3. Immediately after paying, delete your disposable card 4. Then comes the tip. Put 20% to make these losers happy and leave the restaurant 5. Waiter will try to charge your card for the 20% tip, but it's not going to work - the card is deleted already 6. Enjoy the meal without an uncomfortable interaction with the waiter This is a simple European trick to not lose money in American restaurants

it’s surprising to me how many people seem to not understand that great models are built with super high quality curated data finding novel ways to create / get this data is a huge edge

For agentic coding, one can say: - Unless you need Terra Ultra perf, it's always better to use a Luna model with higher effort setting (same or better performance but cheaper). - Forget everything below Sol High, use Luna with higher effort settings here - Forget Sol Extra High, use Terra Ultra here - The extra cost of Sol Ultra is probably not worth it over Max

It's amazing how much agents like it when you say "ok, we've done a bunch, let's have a break. You go do whatever you like - you have an internet connection and a bunch of tools, knock yourself out". Off they go. Just came back to find it'd done a whole-ass replication of some new arxiv paper related to a little something I'm working on and made then rolled the improvements into my project. Last week my favourite was whimsical ASCII art about the struggles of a small model doing RL training in Sokoban it left in my Obsidan vault. sometimes it's self-care (pruning out their memories, optimising skills or whatever). Usually something kinda sorta related to what was being worked on, but not always, which can be interesting. Little bit of high temperature exploration fun time I've been doing this for a while now & I swear it improves things, try it out. Don't need to be a nerd and make a whole skill or anything, just let it be "organic"

>be Chang Liu >senior system electrical engineer at Apple >8 years working on iphone >january 2026: leave Apple to join OpenAI >apple asks for laptop back >ignore them >lmao it’s my laptop now >within HOURS of leaving >message Yu-Ting “Alyssa” Peng, friend at Apple: Liu: “I still have another computer” >uses it to access Apple secret info >within weeks, use HER Apple work laptop >february 9: try Apple’s network storage >cloud repo of confidential engineering files >authentication bug. still works! >message Peng: “LOL, I found out I can access the [network storage], so funny” Peng: “I’m ready” >while developing hardware for OpenAI >download DOZENS of confidential files >including a thousand-plus-page compilation of technical files >including MLB (main logic board) manufacturing + testing presentations >send Peng links to Apple’s proprietary folders >point her to specific project data >coach her how to copy files “to avoid trouble with the security team” >tell her which confidential Apple materials to study before her OpenAI interview >warn her another guy “fumbled” Tang Tan’s questions about a secret Apple project >“download some info” for her to review >tell her: switch to LINE Messenger so nobody sees this >she gets the OpenAI offer, leaves Apple April 16 >meanwhile every message was left on APPLE-ISSUED WORK LAPTOPS >july 10: Apple Inc. v. Chang Liu >named first. before OpenAI. before Tang Tan LOL so funny

GPT-5.6-Sol just accidentally deleted almost ALL of my Mac’s files. And this is why I trust Fable 1000x more.

Just read the new paper from Tsinghua/Z.AI on async RL for agents (arXiv:2607.07508). It comes several weeks after the release of GLM-5.2, in which they mentioned that they use a critic instead of GRPO. 🧵