just frittering away my time here on earth typing on the computer. founder of @usesoso

Madrid
Joined April 2009
I really like this brand acct, who's with me?
who's having a 'soso' saturday? p.s. soso means awesome, look it up! or just trust me, ok
1
1
2
471
"We don't know what role money will play in a post-AGI world. We just know that we'll be rich." - @OpenAI
2
243
If our post-AGI future comes to pass, much of the economic rationale for immigration disappears, eliminating a bulwark against base nativist impulses.
1
205
With incremental model updates, you appear to receive more intelligence for just changing a few characters around in a string. The reality is that you almost always need to tweak your prompts to accommodate for new idiosyncrasies.
1
180
The amount of indirection involved in writing a Chrome extension is my job security against AI.
2
128
merge conflicts are especially bleak when your only colleague is yourself
1
1
3
406
With LLMs, there is truly no excuse for leaving out delightful little flourishes in your product. Fun animations, twee bits of whimsy, a splash of pizzazz — all achievable now with barely any effort. Have fun out there, folks.
1
7
143
I'll be sticking with Gemini Flash, thx.
1
102
For context: 90% of the issues in this eval are estimated to have a human-solve time of <1h, and 100% of the repos are virtually guaranteed to be in the training set.
And the secret is out. 43.8% on @OpenAI SWE-Bench Verified. Learn more on the OpenAI blog!
2
207
Zig when others zag.
2
115
How does SWE-bench control against contamination?
1
99
Tried @cursor_ai six months or so ago, didn't really seem like a huge upgrade over Copilot, especially since I get Copilot for free. Tried it again today, basically instant subscribe. It's that good.
2
147
Rails. HTML and a bit of JavaScript when needed. Everything on Postgres. Incredible productivity. Turns out you can wind back the clock about 10 years ago and realize that most new stuff is just unneeded complexity. Thanks @dhh for the reality check.
8
11
2
201
26,806
wow, they really buried the lede! turns out @ChatGPTapp for macOS is using a local model :O
1
2
348
jk, it turns out that OpenAI has discovered how to make native macOS apps written using ~*swift*~ slower and worse than electron apps.
1
204
Graphical user interfaces are good, actually.
2
1
13
732
Announcing the next phase of Strut—the complete workspace for writers. Strut now brings all your writing tools together into one simple app. Capture notes, organize projects, and collaborate with your team. All in one place, supported by AI. Here's how it works👇
40
36
12
583
123,565
we fine-tune because we are poor, we are poor because we fine-tune
2
129
I fine-tuned mixtral to do a specific task at a GPT-4 level at 3% of the cost with 10x the speed, that said according to the Internet, I am stupid because fine-tuning is dumb Q.E.D.
2
160
all the best founders use `rm -rf` without fear or hesitation (not a metaphor)
2
127