theptip

13 ms·

241.
by theptip 6mo ago
11% further along the particular bell curve of SWE-bench. Not really easy to extrapolate to real world, especially given that eg the Chinese models tend to heavily train on the benchmarks. But a 10% bump with the same model should equate to
242.
by theptip 6mo ago
It’s a good thing to keep in mind, but LLM + scaffolding is clearly superior. So if you just use vanilla LLMs you will always be behind. I think the important thing is to avoid over-optimizing. Your scaffold, not avoid building one altogeth
243.
by theptip 6mo ago
It’s $20k for all the vulns found in the sweep, not just that one. And last security audit I paid for (on a smaller codebase than OpenBSD) was substantially more than $20k, so it’s cheaper than the going price for this quality of audit.
244.
by theptip 6mo ago
I think them stating this very simple policy should also be read as them explicitly not making a more restrictive policy, as some kernel maintainers were proposing.
245.
by theptip 6mo ago
Thats not what a moat means in business. It doesn’t mean impregnable, it just means expensive or difficult to cross. It is absolutely a moat if only $1T companies can afford the capex to compete.
246.
by theptip 6mo ago
Public APIs get distilled, this is why Deepseek and Qwen are so competitive. I am very confident that frontier models won’t be public at strong AGI levels, and certainly not at superhuman levels.
247.
by theptip 6mo ago
Yeah, new feature dropped a couple weeks ago. https://code.claude.com/docs/en/scheduled-tasks
248.
by theptip 6mo ago
I’ve been playing with /loop 5m check if you have any actionable tasks for this scenario.
249.
by theptip 6mo ago
IIUC the operating theory is that a short burst of acute inflammatory stimulus clears out the system to below the prior baseline.
250.
by theptip 7mo ago
They will certainly make human soldiers unviable. (I draw mostly dystopian conclusions from that prediction.)
251.
by theptip 7mo ago
I’m missing something here. Isn’t the best “doing” to actually use Claude to build stuff? The barrier to entry is so low. Why do you need to memorize slash commands? They are somewhat useful and you can just read them from the autocomplete.
252.
by theptip 7mo ago
I tried to put together something similar… I guess it is reassuring that others are struggling to find a nice e2e setup as well. I was intending to vibe code the whole pipeline then stumbled onto Readwise, $10/mo is currently cheap eno
253.
by theptip 7mo ago
Bayesian reasoning is just another algorithm for predicting from experience (aka your prior). I took the GP to be making a general point about the power of “next x prediction” rather than the algorithm a human would run when you say they ar
254.
by theptip 7mo ago
Exactly. Lots can be explained just with more abstract predictors, plus some mechanisms for stochastic rollout and memory.
255.
by theptip 7mo ago
Nice! I was just implementing this for CA state bills. Is the parsing/uploading code shared somewhere else? Definitely the kind of idea that would have been below my activation energy pre-Claude. I think this approach should be standar
256.
by theptip 7mo ago
Yes! What do we want technology to do for us? In my opinion one important promise that was never fully realized is helping us to live a more enriched life. Social media does help people stay connected but it brings a lot of negatives that a
257.
by theptip 7mo ago
In this toy model, the strategy that (may) outcompete is delivering a better service. > The end-game is always monopolization This is one happy end state of PE but let’s be clear that for large swathes of human history it has not been th
258.
by theptip 7mo ago
I’m quite bullish on disintermediating the algorithms. AI makes it very easy to plug in your own. We just haven’t figured out the plumbing yet. I’d be strongly in favor of interoperability laws to pry open the monopolies. (One dynamic you d
259.
by theptip 7mo ago
I dunno, careful what you ban; TV has “infinite scroll” too.
260.
by theptip 7mo ago
I think there is the potential for a beautiful correction here; the general narrative over the last 5-10 years is PE buying up good businesses, enshittifying them and loading up debt to improve margins, and driving down quality. But this le
261.
by theptip 7mo ago
For re-running, I assume you want to do this all on a review app with a snapshot of the DB, so you start with a clean app state. Should be pretty easy to make it deterministic if you follow that precondition. (How I had my review apps wired
262.
by theptip 7mo ago
Ok, here is the risk of being left behind - if we have moderately fast take-off, the 1-2 years required to upskill in AI might mean you find yourself unemployable when your role gets axed. I don’t think folks are taking seriously the possib
263.
by theptip 7mo ago
I agree uv is great but let’s not get carried away here. Poetry is good, pip was fine for many use-cases after they added native lock files.
264.
by theptip 7mo ago
It’s interesting, LeCun seems to have a blind spot around in-context learning. I didn’t find one mention in this paper (only skimmed the full paper so far so may have missed), which is odd as it is the way that agents come closest to autono
265.
by theptip 7mo ago
Is that true though? I trust MSFT, AAPL, GOOG, META more than I trust every website that would be collecting IDs. Most of those guys already have my payment info anyway. Centralization at the major corporations allows (in principle, if regu
266.
by theptip 7mo ago
Perhaps, or perhaps there are legitimate privacy concerns with requiring every website to collect a photo of your ID to prove you are not a minor?
267.
by theptip 7mo ago
It’s simply about making standard, centralized plugins available. Right now Claude benefits from a “link GitHub Connector” button with a clear manifest of actions. Obviously if the self-modifying, Clawd-native development thing catches on,
268.
by theptip 7mo ago
Here is my take - Old world - each platform writes APIs and then has to publish rich client libraries. Despite the server APIs theoretically being well documented, because users want the quickest time-to-first-demo platforms also ship a bun
269.
by theptip 7mo ago
> the sponsor of Louisiana's HB-570, publicly confirmed that a Meta lobbyist brought the legislative language directly to her. The bill as drafted required only app stores (Apple, Google) to verify user ages. It did not require soci
270.
by theptip 7mo ago
> We've all seen LLMs spit out garbage bugs on the first few tries. I’m assuming here an extrapolation of capabilities where Claude is competitive to the median OSS contributor for the off-the-shelf libraries you’d be comparing with

More ›