Pinned Tweet
Claude haiku 4.5 solves 65% of bug challenges, after optimizing the CLAUDE md file through GEPA, you can achieve rates up to 85% (20% increase from the instructions file alone.
Full post here:
tim.waldin.net/blog/2026-04-…
14
34
657
69,571
tim waldin retweeted
OpenAI: meet dots
everyone: you mean bots? like Grok Bot?
OpenAI: meet ChatGPT Space
everyone: umm… you vibe coded Notion?
OpenAI: meet Decisions API, Jev killer
everyone: so you copy pasted a viral startup from 2 weeks ago?
OpenAI: meet sign in with ChatGPT
everyone: congrats on inventing OAuth
OpenAI: meet GPT-6.1 Sol
everyone: but where's GPT-6.1 Astra???
OpenAI: meet the $500 plan (25x)
everyone: wait. so $200 plan isn't 20x?
OpenAI: did you see Tibo's reset button?
everyone: but you owed us one anyway
OpenAI: here's a Chromatic GameBoy
everyone: …
just to recap OpenAI’s DevDay…
- GPT 6.1 SOL is worse than Opus 5.5
- dots is their competitive answer to Grok Bot and Muse (shall see how good, it botched its own demo)
- your $200 plan is no longer a 20x
- a $500 plan is now a 25x
now plz explain to me why Tibo was shouting DevDay for weeks straight
insane amounts of mental gymnastics and arrogance going on at OpenAI just to call today a win
41
44
644
113,455
computer use becomes 100x better when u give the agent their own virtual screen with betterdisplaycli
2
43
tim waldin retweeted
I feel the same way about multi-monitor setups
just use a good window manager with keybinds
Having more than one thread open at the same time is the ultimate LARP.
I regularly have 40+ threads running concurrently. I've never had a reason to look at more than one at a time.
12
2
36
3,429
tim waldin retweeted
playing with jev has taught me that if you make a hammer cheap enough, everything will start looking like a nail
40
166
3,992
122,291
tim waldin retweeted
you either die a coper or live long enough to see yourself become a larper
5
2
71
1,424
just got access to JEV and replaced my prefrontal cortex with it, I’m finally free from the burden of choice
thank you @typesafeai
8
23
388
7,145
tim waldin retweeted
Announcing Lindy Brain: a self-growing wiki Lindy updates based on what it learns about you and your business.
Imagine an employee that's sat on every meeting and read every email and Slack message. That guy's gonna have a good sense of what's going on!
Lindy organizes its learnings across people, projects, decisions and preferences. It's like the self-updating brain of your business, all stored in .md files in the Files section of your Lindy workspace.
Available today! lindy.ai
13
6
62
7,443
tim waldin retweeted
don't worry guys i got this whole ai safety thing figured out
77
26
2,560
238,184
tim waldin retweeted
The year is 1997. The President was born in 1946.
The year is 2007. The President was born in 1946.
The year is 2017. The President was born in 1946.
The year is 2027. The President was born in 1946.
507
16,066
299,880
7,485,179
tim waldin retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
20,637
167,480
802,917
174,715,357
tim waldin retweeted
a session you compacted 157 times is like a seasoned cast iron pan
87
252
6,072
170,967
tim waldin retweeted
openai:
"yeah so we solved math"
me (a genius):
"haha i'm navier strokin my shit rn"
7
3
199
4,227
