Computer Toucher

111 posts

Computer Toucher banner
Computer Toucher

Computer Toucher

@computerusr

autocomplete supervisor sharing lessons learnt

United States Se unió Temmuz 2025
606 Siguiendo155 Seguidores
Tweet fijado
Computer Toucher
Computer Toucher@computerusr·
Why are they playing philosopher in my coding machine
English
0
0
7
710
Computer Toucher retuiteado
staysaasy
staysaasy@staysaasy·
The rate of people showing me their new Replit/Lovable/vibe-coded apps has dropped PRECIPITOUSLY in the last two months.
English
51
16
758
54.3K
Gary Varner
Gary Varner@GSVarner·
@thdxr Those side chats with coworkers are the best especially when they are then made real by a billion dollar company.
English
1
0
11
1.7K
dax
dax@thdxr·
pretty much every competitor in our space has been very easy to deal with except openai, they're the only company that understands building things for a lot of people we basically have no shot at directly competing
English
108
39
4.3K
322.9K
Computer Toucher
Computer Toucher@computerusr·
@knowclarified It feels like we’re going backwards suddenly. These labs just need to put the weights in the bag. They’re not ux designers
English
0
0
3
115
Chris
Chris@knowclarified·
I fucking hate how Claude code puts everything in a stupid worktree what is even the point just edit the repo
English
10
0
51
6.8K
Computer Toucher
Computer Toucher@computerusr·
Gpt-5.5 has this absolutely infuriating trait where it tries to confuse you in the hopes that you’ll get tired of reading its walls of text and stop asking it to do things
English
0
0
3
52
Computer Toucher
Computer Toucher@computerusr·
@kunchenguid Not sure! That’s your job lol Maybe drawing inspiration from a game like territorial (dot) io, where the strategy is based off the color of the terrain? It’s a hard problem. But a benchmark based on elo is much more fun than a % on a test
English
0
0
1
41
Kun Chen
Kun Chen@kunchenguid·
@computerusr cool idea! how would you gamify something like visual design tho?
English
1
0
1
89
Kun Chen
Kun Chen@kunchenguid·
LLM benchmarks are boring. introducing - Trial by Combat!!! two LLMs walk into a turn based strategy game, only one walks out GPT 5.5 completely demolished Opus 4.7 - much faster turns, lower token usage, and it won, fair an square details shared in thread below 👇
English
7
1
71
6.7K
Kun Chen
Kun Chen@kunchenguid·
@thdxr actually, maybe only anthropic is confused
English
1
0
4
551
dax
dax@thdxr·
the whole "cannot use our models to develop a competing product" is getting really confusing now that model labs are launching regular products like Claude Design
English
72
56
2K
327.2K
Computer Toucher
Computer Toucher@computerusr·
This guy is a great follow
Kun Chen@kunchenguid

@thdxr these LLM companies seem confused about what they are they are the power plants we want them to make power cleaner, cheaper, and more reliable instead, they keep building dishwashers and refrigerators

English
0
0
3
66
Computer Toucher
Computer Toucher@computerusr·
GPT-5.5 is the weirdest model I’ve ever used and it’s not even close. It’s a freak
English
0
0
1
48
Computer Toucher
Computer Toucher@computerusr·
Why is every model completely unaware of what it is? “What model are you?” GPT-5.5: “I am chat gpt 5.1!” Just add information about it to the system prompt. Or even an invisible skill to use when asked about this stuff. Am I missing something?
English
1
0
2
77
Computer Toucher retuiteado
Kun Chen
Kun Chen@kunchenguid·
a few really interesting things are manifesting at the same time - 1. git is already decentralized protocol. it’s extremely easy for anyone to self host a git server without needing a “hub” 2. pull requests are dying because of AI. for open source, a prompt that has high quality requirements is better than a PR that no one has time to review. for businesses, review time is shifting left to happen between agent and human author, less and less between human author and human peer 3. CI/CD is a complete commodity - too many solutions available that can do the job better than github actions 4. copilot went from the SOTA to a commodity within just 3 years. they have no edge at all against frontier labs so if storage can be decentralized, PR is dying anyway, CI/CD is not defensible, AI is a losing battle, then what’s left for Github? it’s mainly the community and network effect, which is now also falling apart because of the poor service quality this is NOT looking good…
English
2
2
39
1.9K
Gergely Orosz
Gergely Orosz@GergelyOrosz·
WOW. Mitchell Hashimoto voting with his feet: Ghostty is leaving GitHub. "I can't code with GitHub anymore. I'm sorry. After 18 years, I've got to go. I'd love to come back one day, but this will have to be predicated on real results and improvements, not words and promises."
Mitchell Hashimoto@mitchellh

Ghostty is leaving GitHub. I'm GitHub user 1299, joined Feb 2008. I've visited GitHub almost every single day for over 18 years. It's never been a question for me where I'd put my projects: always GitHub. I'm super sad to say this, but its time to go. mitchellh.com/writing/ghostt…

English
32
33
1.2K
152.3K