oefrha

Apparently open source models is an officially stated national strategy now, which is somewhat reassuring that we'll probably keep getting these for a while.

https://www.reuters.com/world/asia-pacific/chinas-xi-promote...

There’s reason I wrote high info density documentation and comments rather than committed every StackOverflow page I viewed during development pre-AI.

I wouldn’t call it a recipe for disaster, but oh boy if you leave an agent that “hammers until it gets there” on its own with an underlying bug in a dependency…

Yeah, I checked usage stats and pretty sure quota consumption on Max plan is not linear wrt to usage by API pricing. Fable burns quota faster than 2x Opus with equal token count.

Plus I'm also not super impressed; it somehow managed to implement a 200L custom TCP server for a simple static HTTP mock server for a single test case (all that was needed was a fixed route returning a fixed placeholder string) just yesterday. Never seen anything like that.

I have this in my global CLAUDE.md after being annoyed by all the random crap memories.

Don't start generating an auto-memory entry before asking me. Ask first, write only if I confirm — no speculative drafting.

No more crap after this.

Incidentally I don’t recall Opus 4.8 asking me once in the past few weeks. Older models did ask semi-frequently.

Which is a good thing. Self-serving motives are more reliable than altruistic ones.

When the biggest thieves are on track to trillion dollar valuations, what do you expect. Everything on the Internet is free for all now, don’t kid yourself.

That’s just a wall of text for “malware detection is hard, write everything yourself, don’t use third party”. Thanks for the insight, I guess.

anthropic red lines

Alleged red lines. Could be just talking points for garnering sympathy. Big tech aren’t exactly known for being truthful, especially big tech partnering with esteemed Palantir.

Hmm I have a couple Alibaba Cloud servers located in CN Hangzhou region and I just pay with my U.S. credit card, no Alipay needed. alibabacloud.com is meant for international customers anyway. I’m not sure if aliyun.com accepts international CCs, maybe you’re using that.

To save you a read of the original while still having enough context, here's the whole section on crypto:

crypto: spec and test-driven development to the rescue

Cryptographic code is famously hard, with many, many footguns haunting unsuspecting developers (and even experts!).

But, cryptography also has something that you likely won't find in any other domain: an extensive public collection of test vectors, particularly for edge cases. Every algorithm specification come with a basic suite of test vectors, but there are also community-built wonders such as Wycheproof.

These test vectors, combined with the official specification documents of the crypto algorithms were rather effective to guide the coding agents and avoid the worst hallucinations.

Cost: ~ $30

Time: multiple days of part-time work.

I'm under no illusion that the crypto crate is currently bug-free, but if my experience told me anything, it's that even experienced programmers are shipping bugs in crypto libraries. So, for similar outcomes, but using 1/100 of the time and at maybe 1/1000 of the cost, I think it's a clear demonstratin of AI's effectiveness.

Yeah, terrifyingly clueless, don't use.

Sounds good? They can pay for code they want to train on. There are plenty of companies sending me offers to code training materials for them for $50-100/hr. Don’t expect to charge me an arm and a leg for inference and then also train on my code.

Call it Intelligence Store and charge… wait for it… 30%.

I wouldn’t mind so much if it didn’t automatically create the user folder based on real name on the MS account, with space in it, fucking up and complicating all sorts of things.

My status line shows 5h and 7d quota usage among other things like model, context usage, git branch, etc. Unfortunately the whole status line disappears half of the time from CC’s broken ass UI.

Sounds reasonable given the recent YOLO GC debacle https://discuss.python.org/t/reverting-the-incremental-gc-in....

Nobody nowadays goes "you are an expert software engineer. make no mistakes"

You know what, I checked Opus 4.8's instructions to a review subagent the other day and it literally opened with

You are a senior infrastructure/security engineer doing a thorough, adversarial code review...

I didn't say anything like that myself.

I’d be shocked if the Pentagon isn’t running an AI propaganda mill targeting <insert any region of the world>.

Apparently https://github.com/scanaislop/aislop with AGENTS.md is not AI slop, but projects of mine last modified before 2016 all contain critical AI slop. So very accurate.

when was the last time you heard about the Chinese space stations outside China?

Last year, when negative news of delayed astronaut return was all over American news, e.g. [1][2]. Apparently makes American astronauts onboard Boeing ship being stuck in space less embarrassing.

[1] https://www.nytimes.com/2025/11/04/science/space/china-space...

[2] https://edition.cnn.com/2025/11/05/china/china-shenzhou-20-a...

Using text-align: justify for questionable aesthetic purpose here really hurts readability, especially on a narrower viewport like the 1026px viewport of Safari with sidebar on an iPad Pro 12.9’’ (although it’s probably more of a problem of the four column layout on that specific narrow viewport; three should be better).

Yeah it’s smarter than that, but as a daily user of git absorb it still gets things wrong fairly often though—like a couple times a week often for me. Plus the changes it can’t absorb automatically (e.g. a lone doc change it can’t find peers for).

I think there’s something very wrong with every single walk of society, and academia’s problems are far from the most grave. If I’m given the choice again, I take a rotten academia over fucking ad machines and quants that do no good / actively do harm in this world (both industries try to hire me and the likes of me) every time, even when I know I’m gonna leave at the end.

Sounds like the bitter words of someone who got pushed out. I know the type, I’m no longer an academic myself. Sorry it didn’t work out for you, not sorry to claim that the overwhelming majority of the most important advancements are still made by people with PhDs, however many unsuccessful ones there are.

If you mean “intelligent” Americans who work on fucking blood sucking ad tech or garbage financial engineering rather than pushing the boundaries of science and technology, sure.

Go to any top STEM PhD program and do a headcount. I don’t know what’s going on now thanks to this wave of xenophobia and funding cut madness, but back when I was in one (Princeton Physics, that was last decade), everywhere I go it was at least 50-50 in terms of international representation. You can also count the massive number of clearly foreign born faculty. It could not be more obvious.

Edit: And before you mention O-1 and friends for highly accomplished individuals (maybe that's not affected for now? Honestly have no idea), this kind of policy has wide ranging second order effects even if it doesn't affect top talent directly. Like I said I was U.S. educated myself, once I would encourage bright minds from elsewhere to pursue a higher education in the U.S., now I heavily advise them from even setting foot in the U.S.

Not really, it this is too be believed

At the post office, a friendly staff member confirmed it could be sent, helped me package it up securely

It's actually at least half a case of not blindly trusting AI, and pressing until a positive answer is given:

I asked ChatGPT how to send the laptop, and it gave me a spiel about finding a reliable freight service or courier.

(Later) I should have listened to ChatGPT.

Wait, how the hell can Australia Post charge you the full AUD 111.60 for a failed shipment when it seems to be the fault of the clerk who approved the shipment against their own rules? And sounds like the package didn’t even leave Australia so even if you should pay for the full mileage it would be 20% at most?

Failed businesses also tend to provide lower returns on capital, and that’s totally backed up by data. Doesn’t mean the “purpose” of those companies was to provide lower returns on capital.

allow ticket resale only through the official platform and cap it at the original sale price.

That obviously doesn’t work because money can still change hands outside the official platform, unless you mean resale to a random buyer selected by the platform, in which case the resale is not terribly different from a refund and restock for any event where scalping is a problem.

You simply can’t stop scalping if you allow resale. Heck, people even attempt to scalp things where there’s no official resale mechanism (e.g. I change my id at this second, you immediately change yours).