AlexErrant

26 ms·

1.
by AlexErrant 6d ago
This doesn't solve grinding? If it did, we'd see results by now (Published: 27 January 2026), and there's months of delay between submission and publication. I'm willing to be wrong, but I'm just not seeing anything
2.
by AlexErrant 6d ago
Agree. Gonna note that biology is even less grindable than ordinary chemistry/physics/engineering. We'll have advancements in other real-world fields before AI-powered bioengineering becomes A Thing.
3.
by AlexErrant 6d ago
Sup fellow Alex. > AI with real-world competence is coming very soon Disagree, Moravec's Paradox remains undefeated. How long has Elon promised self-driving cars? Or how long have we been seeing humanoid-robot-walking demos? Laundry
4.
by AlexErrant 6d ago
Literally 2 days ago, from a former Anthropic researcher: > The people building AI earnestly believe that it could kill us all by the end of the decade. This is not a marketing stunt. https://x.com/hilbertspaess/stat
5.
by AlexErrant 6d ago
My main issue with this timeline is that AI still has trouble transitioning to the real world. It predicts for 2029: > There are swarms of insect-sized drones that can poison human infantry before they are even noticed; flocks of bird-si
6.
by AlexErrant 8d ago
Ah derp; that's what I get for moving too fast. Genuine thanks for calling me out on my bullshit. (and this is also why I prefer auditable citations instead of casual "my reading of twitter is...") I retract the projectnash c
7.
by AlexErrant 8d ago
2. Fair, and further I would agree that OpenAI not knowing if prior user prompts were part of training data is concerning and will only lead to more secrecy. 3. ctrl-f "x.com" in this thread only yields https://x.com&#x
8.
by AlexErrant 8d ago
This clarification substantially changes the flavor/nuance of your OP; may I suggest an edit (assuming the locktime hasn't passed)? Separately, I disagree that intellectual work has ever been free of "dark forest"-style
9.
by AlexErrant 8d ago
1. What does the dark forest have to do with this? Because "the most senior OpenAI researchers" are shitposting on social media, we've an answer to the Fermi paradox??? 2. The dark forest is fun for scifi stories, but is math
10.
by AlexErrant 13d ago
Cookie nonsense aside, the EU is mandating that AI companies watermark their output. I'm curious if that (noticably) diminishes the quality of the output.
11.
by AlexErrant 13d ago
Sebastian Raschka, author of Build a Reasoning Model (From Scratch) , commented on this: > Reusing layers does not by itself suppress visible chain of thought. It adds computation in hidden states before the next token is emitted, just
12.
by AlexErrant 14d ago
So... is it? You seem to be implying the opposite.
13.
by AlexErrant 16d ago
You're missing the pattern of behavior from Anthropic. They constantly market like this: the session link, the Co-Authored by Claude, the unwillingness to adopt AGENTS.md. Hell, on workflows on 2.1.246 Claude's been ignoring the &
14.
by AlexErrant 16d ago
> even though I'm using a free tier Erm, that's to be expected. GPUs ain't free. Anyone can make a free account and cost them potentially infinite compute, so a phone number is a way to lock that down. Might not be the &
15.
by AlexErrant 17d ago
If _you_ choose to do it, no. If the tool auto-does it, yes.
16.
by AlexErrant 29d ago
> we’ve also switched to a reprogrammable keyboard controller using open source ZMK firmware o.m.g. Is this the first laptop to do this?
17.
by AlexErrant 29d ago
Opus5 decided it would be a good idea to include my email as part an HTTP request header's User Agent because Anthropic includes your email as part of the system prompt. Fucking morons over there.
18.
by AlexErrant 29d ago
Another week, another outage, another cache expiration of my prompts through no fault of my own. At least OpenAI has the decency to reset after a serious outage.
19.
by AlexErrant 1mo ago
Agree 100%. To quote myself: --- Which do you think is worse: smoking, or not exercising? https://jamanetwork.com/journals/jamanetworkopen/fullarticle... Look at Figure 2C in particular. The hazard ratio of smokin
20.
by AlexErrant 1mo ago
People kick around the idea of the "personal fat threshold". E.g. caucasians seem to be better at carrying fat than, say, Southeast Asians. White people seem to be able to better store fat without metabolic side effects; SEA seems
21.
by AlexErrant 1mo ago
To "actshully" you (because there are no absolutes in biology) pro-triathlete Lionel Sanders got his hba1c up to prediabetic range (5.9) https://www.tri247.com/triathlon-news/elite/lionel-sanders-t... Al
22.
by AlexErrant 1mo ago
Is this performative posting? We're supposed to care that you don't care? Could you imagine this forum if everyone acted like you?
23.
by AlexErrant 1mo ago
To not answer your question, I have a claude stop hook that my status line uses to tell me how close my cache is to expiring https://pastebin.com/JLXUG16Q _entirely_ vibecoded don't @ me.
24.
by AlexErrant 1mo ago
A worthy and pragmatic tradeoff. I've always appreciated Solid's DX/mental model (I _never_ have to think "when does this re-render?") so v2.0's continued wins on DX are remarkable. I never thought a release co
25.
by AlexErrant 1mo ago
I'm a little surprised that there's no mention of end-user perf and that it's not on the js-framework-benchmark. Not an accusation; just noticing the lack of emphasis (perhaps natural, given how much Solid 1.0 was famous for
26.
by AlexErrant 1mo ago
Here's what it sounds like on Android https://www.youtube.com/shorts/HQe8e3voRdw IDK why it's so much more chill than the "child abduction" alerts I get. Those are straight up adrenaline-inducing so
27.
by AlexErrant 1mo ago
This quote is why I have my Claude Code "done" notification sound be the sound of a lawnmower revving. The lawnmower can't hate you. Don't anthropomorphize the lawnmower, no matter how clever it sounds. (Actually not tha
28.
by AlexErrant 1mo ago
Sure, you can "vibe/auto-retro". But in my experience the retro often results in "higher-order" structural changes. In starker terms: if you're ok with "vibe principles", "vibe invariants",
29.
by AlexErrant 1mo ago
I literally hold a "retro" with my agent after a session. > Let's conclude with a retro. Did you run into any issues during this session that you think could be improved? Any failed tool calls, confusing docs/prompts,
30.
by AlexErrant 2mo ago
Even Lean has bugs. > AI "Proves" Collatz Conjecture with Lean 4 Bug https://news.ycombinator.com/item?id=49101465

More ›