Wes Winder

y'all realize you can just put "@AGENTS.md" in your CLAUDE.md right?

claude code projects is just a coordinator agent that spins up parallel threads you describe the work, it splits it, tracks it, stitches it...

gemini 4 argon is #1 on deepswe and automationbench 77.9% and 51.3% but it's still locked behind google's "trusted testers" meanwhile, no...

y'all realize max reasoning is a waste of money right? astra on max costs way more than xhigh and literally scores worse on terminal bench ...

your agent transcripts are the most underrated dataset you own once a week, point an agent at the last 7 days of sessions and ask for: > m...

claude code is deleting your session transcripts by default it only keeps them for 30 days then they are wiped forever but to keep them f...

haiku 5.5 is clocking in at 122 tps on openrouter btw nearly 2x faster than luna https://x.com/weswinder/status/2107939519961997441/photo/1...

anthropic may have just put the final nail in the coffin for luna haiku 5.5 looks ridiculously good crazy to see a claude model getting so...

tibo is scraping the bottom of the barrel already this has nothing to do with codex i think it's time for a reset ya'll https://twitter.co...

best thing anthropic can do right now is to say nothing at all then drop fable 5.5 openai is cooking themselves on the timeline (@poteto i...

wow tibo actually listened they bumped sol 6.1 from 30 tps to 50 tps love seeing openai address the main problems now we just need a mode...

chatgpt is just full of upsell slop now this is on the pro plan btw

it really feels like spacexai is the only lab left that still has compute openai can no longer make cheeky comments about their compute cap...

chatgpt has slowly turned into enterprise slop and it's kinda sad

i just realized i'm basically making a web browser for coding agents lol

i don't read the agent chats anymore i simply talk to one orchestrator agent it does everything under the hood i only need to talk in hig...

i'm sure sol 6.1 is great but i can't handle how slow it is compared to opus 5.5

you can tell how uncreative ai is based on how uncreative the big labs are these labs all have access to ultrafast next gen sota models bu...

this aura loss for openai in the last month is honestly insane it needs to be studied so many back to back to back fumbles smh they’ve co...

i feel like openai is getting shiny object syndrome again after seeing success with codex