sdmat Profile
sdmat

@sdmat123

Followers
339
Following
138K
Media
181
Statuses
6K

AI, shitposting, philosophy. Maybe some systems engineering.

Joined February 2022
Don't wanna be here? Send us removal request.
@sdmat123
sdmat
3 days
RT @deredleritt3r: Let's assume that AI labs can charge as much as Netflix per month (they currently charge more) and that they'll never ha….
0
8
0
@sdmat123
sdmat
16 days
RT @sdmat123: @scaling01 Looking at performance over log price gives a clearer picture. o3 and 2.5 pro are the immense breakthroughs. Both….
0
2
0
@grok
Grok
6 days
Join millions who have switched to Grok.
246
488
4K
@sdmat123
sdmat
18 days
Tweet media one
0
0
7
@sdmat123
sdmat
18 days
RT @edwinarbus: good advice from @__ruiters:. GPT-5 isn’t broken. Your prompts are. A lot of folks, myself included, expected GPT-5 to be….
0
48
0
@sdmat123
sdmat
24 days
RT @pashmerepat: Seeing a lot of conflicting takes about gpt5. If you have bad ideas, taste, or bad comms, the model is too autistic to m….
0
57
0
@sdmat123
sdmat
26 days
For anyone trying GPT-5 in Codex CLI and wanting to set reasoning effort this is how to do it:. codex -c model_reasoning_effort="high".
2
1
10
@sdmat123
sdmat
26 days
RT @VictorTaelin: Nah you're all wrong, GPT-5 is a leap.I'm 100% doubling down here. I didn't want to post too fast and regret it again, bu….
0
47
0
@sdmat123
sdmat
27 days
RT @Teknium1: Starting to feel like this gpt oss was trained on like 20T tokens of distilled safe maybe even benchmaxxed data from o3. Ther….
0
63
0
@sdmat123
sdmat
1 month
RT @sdmat123: @VTchuiev @DannyAziz97 You're not going to make it as a vibe coder if you worry about little details like the code working.
0
1
0
@sdmat123
sdmat
1 month
Using AI for complex projects is to experience all the joys of being a line manager only your reports also have anterograde amnesia.
0
0
3
@sdmat123
sdmat
1 month
Seeing victims everywhere you look is the zeitgeist, it is what gives people license to behave as they like while feeling virtuous. This view of virtue is both sentimental and unfeeling, cloying and brutal: for it implies that those who are not victims are not worthy of our.
3
0
8
@sdmat123
sdmat
2 months
RT @sdmat123: @dissidentwest It's not deep, base models are trained on the internet. And the internet is full of raging racists and antisem….
0
21
0
@sdmat123
sdmat
3 months
For anyone in doubt about the morality of the Iranian regime, consider two details about executions. First: short drop hanging. Civilized countries that hang people carefully set up the noose and drop so that the condemned has their neck sharply fractured and instantly loses.
6
4
13
@sdmat123
sdmat
3 months
RT @sdmat123: @wesyang Just before Kristallnacht, Goebbels gave a speech. He said demonstrations against Jews were not to be prepared or o….
0
12
0
@sdmat123
sdmat
4 months
OpenAI Codex is great, but for a small team / single author project reviewing the pull request in Github is redundant - you just reviewed the change in Codex!. My solution: automatically merge codex PRs with a Github workflow. Just sub in your username: .
Tweet card summary image
github.com
Bring your project into LLM context - tool and MCP server - smat-dev/jinni
1
1
6
@sdmat123
sdmat
4 months
RT @sdmat123: @francoisfleuret When you say "zap you with a taser so that you forget what you just did" that implies a disruption of mental….
0
2
0
@sdmat123
sdmat
4 months
RT @sdmat123: @ArtemisConsort "I don't know anything about this and won't look it up. What you say contradicts my worldview so let's assume….
0
13
0
@sdmat123
sdmat
4 months
RT @sdmat123: @IsaacKing314 "It was a completely unbiased process, we walked around randomly until we got where I wanted to go".
0
2
0
@sdmat123
sdmat
4 months
RT @sdmat123: @ns123abc o3: If you can't do it in 200 lines it's not worth doing. By me, anyway. 2.5: 1000 line component? Coming right up….
0
11
0
@sdmat123
sdmat
5 months
@daniel_mac8 - you might like this!.
1
0
3