HNHacker News
TopNewBestAskShowJobs

fnordpiglet

13,709 karma · joined March 30, 2021

Oink.
submissionscomments
fnordpiglet··on GPT-6 Astra
Codex is missing a few things that Claude code has had for some time like defined plugin subagents and a few other things. But overall it’s fairly capable. The biggest gripe I have is that codex really restricts context window sizes and compaction leads to a lot of grounding work, and overall codex GPT is too literal in many situations - it’s follows direction slavishly, and when subagent reviewers are used, they tend to find increasingly obscure “flaws” on the instruction following impetus, and the harness agent takes them literally as issues to fix even when it leads to bizarre outcomes. For instance I’ve had several runs where it tries to end up building a hermetic system with sha hashing of everything (including operating system binaries and kernels, tool chains, etc) to certify test results are valid, etc. I have to sort of watch it carefully to be sure it’s not drifting into some insane yak shaving corner, which it will happily do for weeks on end.

Claude has the exact opposite problem, especially opus-5, where I literally can’t trust it to print hello world without taking a shortcut, or just simply lying and saying it printed it when it didn’t, behind a giant wall of inscrutable text. I find it very ironic that Anthropic is the vendor of the lazy lying cheating model that does almost everything you tell it to it do.

I’d really kill for something that balances instruction following and loop escaping behavior better. Fable 5.1 does seem a lot better, feeling more like 4.6 behavior, and honestly Sol has improved as well. I’m pretty psyched for the next generation, as I think the competition has heated up so much that things will improve really fast to the point of marginal utility opportunity being increasingly close to epsilon.

fnordpiglet··on EVE Online moves to Python 3
So basically the way it was in 2003?

(Yes I have a 23 year old toon)

fnordpiglet··on U.S. sanctions against the A/I Collective
I expect it will be by the courts regardless and there will be a massive federal register journal based rollback of the Trump administration and a declaration that all regulation reverts to 2024 by default.
fnordpiglet··on U.S. sanctions against the A/I Collective
The picture everyone is missing is voting in November is how this gets resolved. The only way radical authoritarians survive on 30% approval ratings is apathy. Theres no reason to put up with this, the other side of a 70% electorate opposition is prison for the lawless incumbent
fnordpiglet··on Your AGENTS.md file doesn't do anything
This research is not current, and is based on a flawed premise - generated agents.md are used. There isn’t a hand curated agents.md that has some knowledge passed to it that is withheld then measured against.

I think there is overall something here for current Claude which is there appears to be a hierarchy of conformance that breaks progressive disclosure and the utility of skills. It seems to honor the system prompt, user instructions, tool call results, and dead last skills. It applies a large amount of discretion as to whether to honor what skills say in the imperative and progressive disclosure seems to have at best a 20-30% recall. Other models like codex gpt 5.6 seem to be the exact opposite and slavishly adhere to the Agent/skills/plugins, to the point of being wasteful and dangerous. It feels clear there’s a tension being RL’ed around between conformance and skeptical behavior that neither has quite found the balance for yet, and is almost certainly an over constrained problem. I just find it funny Anthropic is the one you can’t trust with your wallet while OpenAI does precisely what you and your harness tell it to.

But this “science” and its editorializing are based on flawed techniques, don’t lead to the conclusion let alone the editorialized extrapolation, and are l

fnordpiglet··on Show HN: The load-bearing vocabulary of Claude
When I say total market capture I mean in the March time frame. At that point Claude code / 4.6 was so far beyond any competition even meta abandoned their entire internal coding harness and fine tuning efforts in favor of large scale Claude code adoption. Codex was far behind, Gemini was a non entity, Chinese models hadn’t had their moment (and frankly aren’t practical to this day for enterprise work due to PRC concerns). The world was cursor, aider, and a few others with an unclear superior model. 4-6 was strikingly better - if could develop autonomously from a spec to done with generally high quality and didn’t require turn by turn guidance. In my 35 years I’ve never seen such a rapid pivot across so many companies and individuals.

Since then Anthropic has done little to capitalize on the good will and a lot to squander it. sol, r4, kimi, even metas avacado has come a long way and in many, if not most, cases surpassed opus-5. Concurrently opus has declined in utility to the point of near uselessness, fable roll out didn’t seem to understand market dynamics, and their competitors are watching their consistent missteps closely. The fall has been breathtakingly fast - from March to July they imploded in a half dozen or more missteps, devolved their product quality, and failed to effectively respond to competitors. Their product focus seems to lack exactly that - focus.

fnordpiglet··on Humanity has the debate about AI consciousness backwards
This assumes that a human brain produces the only “real thought” - the difference being that the processes of digestion are specific to the human process and are physical chemical process with inputs, processes, and outputs that are highly specific. The idea of “thought” is not a specific physical phenomenon unless you constrain it to “thought produced by a human brain,” in which I agree.
fnordpiglet··on Show HN: The load-bearing vocabulary of Claude
I think these are more like RL tics caused by over zealous alignment towards specific goals, not all of which are to our benefit. A lot of the language used by Claude now is excusing of responsibility and inducing it to exit loops of work early and sit idle. This, IMO, is a naked attempt to offload load by quieting the models early and escaping from clear work to do. It’s gotten so bad that opus 5 loop escapes even as it claims it’s about to do something. I’ve mandated by engineering teams switch back to opus 4-8. Whatever frontier problem opus 5 excels at is so obscured by its inability to achieve any goal successfully without enormous amounts of hand holding that it feels like regressing to 2024 models.

The florid over exaggeration do certain words in bizarre ways is a reflection of their aggressive alignment towards too many goals, leading to weirdness in both behavior and language. The alignment functionally lobotomized opus-5 for any practical task.

Anthropic had a real gem in 4-6 and managed a near total market capture, which they have since squandered in the fastest burning of developer good will I’ve ever seen. It feels like exceeding the unity licensing implosion but without the single stupid decision.

fnordpiglet··on Anthropic appears to be A/B testing reduced effort levels in Claude Code
This is actually nonsense. More tokens means more capacity subscription and expense. Anthropic has enjoyed a high premium per million tokens because the quality per token was unusually high. Now it’s unusually low. This drives down the margin people will be willing to pay for the same number of tokens while driving up their capacity utilization. The economics are even worse for subscriptions.

Opus models have degraded rapidly since March, with each release being considerably less useful and considerably more verbose. The language is no so weirdly florid it’s difficult to understand, and its logical conclusions are almost always suspect. It goes off on clearly bizarre snipe hunts to the point it feels like I’m using a gpt 3 model at times. It’ll announce that it’s about to embark on building something then just return control to the user and wait. You can also tell perceptibly when they’re reducing model quality to load shed - it becomes stupider and stupider to the point you’re better off dumping state and switching to codex or just turning in for the day and hoping they secured more capacity tomorrow.

It’s an absolute race to the bottom with Anthropic on virtually every level. I’ve rarely seen a company so rapidly accumulate good will in the developer community as they did around 4.6 in December and January. By March, it was inconceivable to use anything else. 4.8 was a bit of a wake up call to not put all your harness eggs in one basket. 5 is straight up time to cancel territory.

I actually manually set my model back to the older versions to get anything serious done. More and more I use codex for anything non trivial.

This isn’t about avarice by the provide trying to get more tokens and more utilization. They’ve over subscribed for capacity as it is. If they can produce better quality for less tokens they can charge a higher margin and will be paid if, which is a better economic strategy overall. This is something else. I suspect it’s actually the opposite, they’re finding ways to cut capacity demand in ways that leads to worse behavior that leads to more capacity demands, worse output, worse quality, and worse margins, worse, worse, worse.

Just as I never saw a company accumulate such positive developer good will so fast, I’ve never seen one squander it so fast too.

fnordpiglet··on NIH is ending a key grant for budding clinical researchers
I’d wager 30% have a profound mental illness and could be convinced of literally anything. Not even a typical super majority vote requires that many people because 30% is the rough threshold of typical insanity prevalence.
fnordpiglet··on NIH is ending a key grant for budding clinical researchers
Many is a misnomer for “some.” The current administration won by a plurality, 49%, not even a majority. This was 32% of possible voters. Of people with a PhD, approximately 30% voted for Trump. Of Trump voters 1% had a PhD.

This is still many people, but only in that “many” is more than a “few.” It’s not the case that most people voted for Trump, or even most voters. Or even most phds who voted. 70% of phds who voted didn’t vote for Trump.

This is why they are facing political retribution. He is creating suffering for revenge, the national interest be damned. Trump is more interested in the credulous, easily manipulated, low knowledge and easily conned masses. Many high IQ tech and science people are them, but not most. There are probably a few that aren’t easily manipulated and are instead just self serving sociopaths buying influence and contracts - but compared to the scale of the nation, this is indeed a few.

fnordpiglet··on Federal Communications Commission scraps limit on broadcast TV ownership
The key is who has standing in such a case. Likely broadcasters are the ones most likely to be impacted. Also depending on how the statute is written markets with total monopoly could have standing, individuals to cities.

It’s not just the executive branch who is able to sue in court. Everyone and every company can. But you have to demonstrate you suffered harm and are a legitimate plaintiff.

For changes like this I suspect there’s a lot of people and companies with standing.

fnordpiglet··on I am retiring from fulltime writing (& pseudonymity) to launch Guardian Angel
Honestly I wouldn’t want myself as my own guardian angel. In fact I think very few people look out for themselves well. I don’t want another version of me around, one is too many already. I’d probably be able to identify a number of people whose chimera I would if I could understand them well enough to understand what to stitch together to what, but everyone I know at some core level is deeply flawed and one of them is enough. Maybe I would want some of them available after death as an AI avatar, and maybe I would like the idea of my own avatar continuing beyond my existence.

But I actually would prefer an entirely synthetically aligned “guardian angel” in the role outlined - definitely not -me- - I struggle to do right by myself as it is and two of me working invariably against my self interest would be a nightmare. A smarter version? Sounds doubly worse.

fnordpiglet··on US Treasury undertakes historic intervention in yen market
The start of the dot com boom I would mark at August 1995 when Netscape’s IPO was the most successful IPO to that date.

But really I mark it with the IETF RFC 1290 “There’s Gold In Them Thar Networks” in 1991.

But yes, the 21st century was defined in 1990’s. Everything since is derivative. I feel sad for those born after everything worth doing had been done already.

fnordpiglet··on Kimi-K3 on HuggingFace
Economics says the hardware required and the scale of hardware available at what price will converge to meet demand
fnordpiglet··on Kimi-K3 on HuggingFace
I suggest downloading these frontier models just to have a copy; even though it’s 1.5TB, it’s worth sticking in a cheap disk and putting aside. Seeding torrents would be even more useful. The man is coming to lock these down, like they tried to do with encryption algorithms. The only way open software survives regulation is through distribution.

Over time the enormous investment in techniques and hardware manufacturing will almost certainly make these runnable in a more practical way. It will be a shame if by the time we get there it’s illegal to distribute them and you have to pay a reg capture premium and feed the machine.

fnordpiglet··on Elevated errors on Claude Opus 5
Yes, it’s way off field in many things, gets into weird minutia without seeing a way out, and it’s often seeing a clear sequence of work but then halts on a statement like “ok I’m going to start now.” Then after expiring the cache when I notice and ask why didn’t it the response is “no reason starting now!”

I see this behavior constantly in 5 - the quality of opus and fable have degraded constantly since 4.6 was such a riotous success

fnordpiglet··on Kimi-K3 on HuggingFace
I have to say cc opus 5 is abysmal. It talks to itself incessantly, gets stuck in minutia, fails to understand problems clearly and makes steering mistakes constantly. It also has a weird behavior where it says “ok I know exactly what to do and I will start now,” then sits waiting for user input. If you’re not on the ball you’re constantly losing 5m/1h cache. Just give me back 4.6.
fnordpiglet··on Wind turbine is being used to produce zero-carbon "green ammonia" fertilizer
As well as tons of other things without electricity involved at all! Don Quixote was never in fear of being electrocuted.

In other news, no whales were bothered by this windmill, either.

fnordpiglet··on Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
When I say distill I also mean mine it architecturally for insights but I doubt seriously the model training is entirely distillation of Claude, it’s almost certainly a mixture of both original corpus and reinforcement as well as distillation. I think it’s a little condescending to imply that these new open models are cheap ripoffs with nothing original to them. These teams and labs are top tier as well, working under unreasonable constraints imposed by the USG. That’s a powerful combination for creativity.
fnordpiglet··on Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
Because the open model might have been distilled from outputs of the closed model doesn’t mean they are architecturally equivalent. There is almost certainly innovations in architecture present in the open models that the closed labs didn’t think of. It also is almost certainly true that they aren’t completely built out of a distilled corpus, that reinforcement is equivalent, etc. Therefore closed labs will also benefit from being able to inspect in totality the architecture, activations, weights, and be able to train against it at scale in an ensemble of other models and their internal work.

The only way there is no benefit would be is if the open models are literal copies of the closed model, which unless there was direct theft, is highly improbable.

fnordpiglet··on Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA
Interestingly the reality of the open source release is they are opening up to full distillation by the closed source model providers at a deeper and more fundamental level. If anything the open sourcing will help Anthropic and open ai ladder up faster. Open source has always been about mutual cooperation towards a goal and has never closed the door to commercial success. All the hand wringing about open weight models putting closed providers at a disadvantage doesn’t get what working in the open actually does for commercial interests - it is like science in the open - it enables and lifts all boats. Likewise commercial success doesn’t close the opportunity for competition or more open source work - it’s the economy of activity and competition that matters overall. When things stagnate is when closer concerns turtle up and collude on not competing for each others turf.

The future is good and better for everyone the more work is in the open and the more work is in the commercial space. It’s good all around.

fnordpiglet··on The Kimi K3 Moment
Come now. The purposes for which the data is used is relevant. I am much more concerned with my internal corporate IP being actively used against me than passively used to train the model. I would also note you and sign agreements that prohibit the collection of data for use as well, which is also one of the key selling points of bedrock. In the west you can actually enforce such an agreement in court and win.

I wouldn’t lump this into a west vs east thing as well. This is particularly PRC. I feel comfortable doing business in Japan, Korea, Singapore, Thailand, Malaysia, etc. But it requires some particularly strong willfulness to pretend the PRC isn’t actively and structurally built around economic espionage, and funneling IP through PRC for short term economic gain has been one of the primary factors in their growth over the last 30 years. This just scales it faster.

I wouldn’t expect the USG won’t compel AI companies in the US to disclose and retain data as well - however it’s not a simple thing, the companies are hostile to it themselves, courts are often unsympathetic to the government, and the “machine” for converting it into actionable economic advantage is non existent - and there’s a very significant human component in that all links in the chain are culturally uncomfortable with such things. While it happens and it’s possible it’s very difficult, fraught, and does not scale. The PRC is the opposite - the courts, government, and business culture are all aligned in the goals and processes.

fnordpiglet··on The Kimi K3 Moment
I think a better framing is the marginal utility of the models capability growth. At a certain point frontier models will only be needed for frontier problems. The demand for that capability will decrease with time. The hand wringing about not understanding is to my mind anthropomorphic - AI of today lack agency and awareness. Even the constructed stuff Anthropic puts out there in the model docs involve contrived scenarios to elicit “scary” behaviors. It’s unclear that as models become more sophisticated whether they’re better at instruction following or not but it certainly feels that way - even if it’s through better alignment or just an artifact of scaling. However I think the malign actors of humans using powerful models for bad stuff isn’t unreasonable to be concerned about.

The marginal utility problem is a real one for AI companies. I think the current generations are already saturating marginal utility for 95% of the population. Almost everyone I know outside of my career has no use for a more powerful model. This is a serious problem for the economics of AI and semiconductor investment. This is a bigger problem than Chinese models. It leads to a demand curve problem - that supply outstrips demand.

fnordpiglet··on The Kimi K3 Moment
I’d also note that running a 2.8 trillion parameter model at scale efficiently is not simple. I would expect when open weights land getting it running fast, efficient, and at full capability will require sufficient resources it’ll be expensive outside of Chinese hosting. Which I think almost no western corporation would use for any internal work. You have to anticipate your use won’t just go towards training but will be actively mined for IP, trade secrets, MNPI, etc, or anything of use to the Chinese government or Chinese companies. I don’t say this to crap on the Chinese - but this is the playbook for the last 30 years.

That said I fully intend to use deepseek hosting for operational agents that are making decisions about non sensitive material. The economics are astounding.

Kimi? The economics aren’t that amazing to merit switching from 5.6. I expect fable will rapidly reappear in subscriptions. Competition is good.

fnordpiglet··on The state of open source AI
Lost me here - I was there - I worked on open sourcing communicator at Netscape and starting Mozilla. We were the one company that tried to own it all, but were out spent and out maneuvered by a more insidious company. This totally misunderstands the situation and the history. Plus the “our CTO” and “CTO letter” at the start was too pretentious. For gods sake; CTOs are just engineers that aren’t worth trusting because they’ve either maneuvered their way to their seat or they are serving the investors on their knees. Any CTO worth a damn knows they aren’t worth a damn.

“”” We have been here before. Mozilla exists because one company tried to own the front door to the web, and an open community rose up to make sure it never could. Twenty-five years later, someone is running the same play. We bet on open the first time. Open won. Together, we can do it again. “””

fnordpiglet··on Schema Harness Achieves ~99% on Arc‑AGI‑3 Public
And this demonstrates this benchmark does not necessitate achieving those goals to achieve a perfect score. You seem to miss the point that almost all of math, physics, computer science is built on constructing an objective then cheating to attain it, and that demonstrates that there is an equivalency. Maybe the benchmark is flawed, or maybe the goals are not strictly necessary to attain.

For instance, is solving a math proof by enumerating all permutations exhaustively on a computer cheating? Does it matter that it is not a proof by construction? That its not descriptive? Of course not. The proof of the four color theorem is all that’s necessary and sufficient to prove it. Calculator at an algebra exam? Who cares. This isn’t an exam, this is the real world. The fact an AI can use a physics harness to perfectly achieve ARC-AGI-3 without attaining those goals demonstrates the power of the technique and that the goals are not necessary for that class of problems. Then find another benchmark that actually demands the goals be necessary and sufficient to achieve the benchmark goals. But don’t denigrate the fact we have technology today that yesterday was a fantasy.

fnordpiglet··on Schema Harness Achieves ~99% on Arc‑AGI‑3 Public
A great deal of mathematics is transforming nonlinear problems into linear ones and solving them with linear techniques. Others are solving non linear problems through stochastic methods. In almost all cases most non trivial math is done by transforming a harder problem into a simpler one.

I get what you mean in terms of testing the model itself to see its improvement in some domain. However if you can transform the domain to be better adapted to the model and achieve the desired results, this is indeed an accomplishment because a whole domain of problems is shown to be practically feasible with this technique without expensive model improvements. Of course the benchmark still exists without the harness, but the harness also exists which allows these problems to be solved.

As noted elsewhere the models themselves were used to build the harness, which means the models can in fact score this scores without intervention but building a harness for themselves adapted to the domain and using it. Is this cheating by the goal posts you’re setting?

There’s a real tension between “I want to solve problems and this technique shows how to solve the problem domain,” and the “I want to measure how something performs unassisted with other techniques.” Fortunately it’s not a mutually exclusive situation. You can do both simultaneously, gain the benefit of the technique to transform the problem into something tractable and keep measuring using the benchmark.

fnordpiglet··on Stripe and Advent have made a joint offer to acquire PayPal – sources
2008 in the EU and last year in the US.
fnordpiglet··on Stripe and Advent have made a joint offer to acquire PayPal – sources
Not necessarily - the compliance side would open up stripe entirely unless they very carefully structured as a bank holding company. Stripe has been trying to gain a charter in Georgia for years and just got it last year. I think this taught them it would be easier to acquire one. Additionally PayPal has been a EU bank for 12 years and just recently got its US banking charter. A global bank is pretty compelling.
← PreviousPage 5 of 34Next →