I'm confused by your messages linked by DJB. You say that better cryptographers would not choose hybrids, which seems to say that you should indeed think that hybrids are not a good choice. Then you say you are not such a good cryptographer and would choose a hybrid. But if you know that more senior cryptographers think they are not the right choice then why choose them anyways? Or am I misreading "cryptography-literate" here?

Can you explain a bit more regarding your statement that DJB's POV on the matter has no broad support amongst his peers? I'm not in the field but Bernstein seemed like a highly respected member with a long track record in the crypto community, at least from the outside. Do you think the community is wrong or is it DJB who's wrong and why? There's also a good chance that I totally missed the argument being made.

I can think of several academic cryptographers (or rather, practitioners with doctorates) who think the hybrid/pure thing is silly. I didn't claim that a majority oppose hybrids. The point of the message he snipped from the HN thread is that pure MLKEM is not considered an unserious design by actual cryptographers (people on the HN thread --- generally not cryptographers, like me --- think it is). Cryptography engineers tend to default to hybrids.

Downthread we develop more clarity about what it is Bernstein is actually in an argument about. It isn't hybrids vs. pure!

I can think of academic practitioners with doctorates who had wildly disastrous opinions about their field of expertise too:

Peter Duesberg, a Ph.D. in molecular biology / retrovirology who taught at Berkley and was a member of the National Academy, who maintained that HIV does not cause AIDS and that antiretroviral drugs do more harm than good.

Walter Freeman, M.D. academic neurologist and first chair of neurology at George Washington University who believed that severing frontal-lobe connections could stabilize personality and stop pathological cycles of thought.

Charles B. Davenport, Ph.D. in biology, geneticist and founder of the Eugenics Record Office.

Henry H. Goddard, Ph.D. in psychology, intelligence-testing researcher and later professor of abnormal psychology who believed intellectual disability, poverty, prostitution, and criminality constituted a hereditary family type.

Clarence Cook Little, Harvard Ph.D., mammalian geneticist and prominent cancer researcher who insisted for years that the evidence did not establish a causal relationship between smoking and lung cancer.

Fun game. We could laugh at all of them, and your examples too, if not for the damage they caused.

3.5 Pro was supposed to be around the corner two months ago. 4.0 Pro is some ways out as they recently stated they are seeing some promising early results from training. It didn't sound like a release is imminent.

Grok, Meta, Gemini and others all released updates to their models within around a month or two from their respective last release and made significant jumps in benchmarks all around the same time. Any guesses as to why that is? Is it just the release season and/or everyone is benchmaxxing?

its essentially the same model being trained continuously 24/7 with the company periodically publishing just a new checkpoint

each new checkpoint can benefit from better reasoning training, RL on specific tasks and more synthetic data

So why do they seem to release around the same time ? my guess is because they time major releases around quarterly earnings, investor meetings and other important business milestones. Once one company announces a major update, the others also have an incentive to ship their latest checkpoint rather than look like they r falling behind.

Sure, they are just checkpoints, that much I guess is obvious. The question is why did they not do frequent releases like this before and why are they making significant jumps in benchmarks so fast and all these companies suddenly falling into that pattern? Earning reports are not to come until end of October, that's not it.

possibly the beginning of the recursive feedback as models begin to aid in their own improvement? especially algorithmic improvements, which seems to have a lot of wide open space for gains

You post your benchmark on every other AI article, I've seen you do this by now more than a dozen times. It's a bit much. I don't want to be too harsh but your benchmark is obviously flawed when the top 3 models for Typescript (Combined) are Grok 4.5, Muse Spark 1.1 (lol), Gemini 3.5! Flash and then followed by Luna, beating Opus 5, Fable, 5.6 Sol etc by quite some margin. In fact 5.6 Sol ranks lower than Kimi K2.7 Code and even Grok Build 0.1. There are so many entries in your rankings that don't make any sense whatsoever that I can't take this benchmark serious and I have not seen it gaining traction. Please stop spamming it?

The reality is that cost is the primary constraint for the public benchmark we provide. While we run enough samples to get results that are generally quite accurate on average, we only produce ~10 coding submissions per language for each model and those are across random environments, which naturally has noise. Plus that's split between agentic coding sessions and one-shot coding.

So just adding a language or tag filter can result in some pretty small sample sizes. You can see how many samples survived in the box plot, but that's probably bad UX that most people never see. There's a reason no other benchmark provides this type of data (even for our sample sizes it runs almost 10K USD/month to keep up to date with new releases).

Might be a good idea to reduce the ability to apply filters into a cohort with less than ~20 samples -- not the first time we've gotten that feedback. Seems like adding too many options to see individual sample variation is just misdirecting. I'm a nerd who loves data so I hate removing access, especially since the aggregate performance is very interesting (averaged across all languages, we see consistent and interesting performance data across models, like models outperforming with strongly typed languages). But tbh I think you're right and we'll try limiting filters to where we actually have statistically significant data.

Google wanted to release 3.5 Pro last month but because of the trouble Anthropic got with Fable they might have wanted to wait a bit for the dust to settle I could imagine. And now there is quite some competition. 3.5 Flash for me is a replacement to 3.1 Pro. It's more like a 3.2 Pro. It costs about the same (or more!) than 3.1 Pro, is a little bit smarter in many cases and a little bit faster. 3.5 Pro will be a lot more expensive and I expect it to juuuust be able to hang with Opus 4.8 and GPT-5.5.

I wish Google was able to actually push the industry further, either in terms of quality (intelligence) or quantity (price) but they've been playing catch up a lot.

They are playing the game a bit differently than all the others. The others have useable IDEs etc. while Google has a boatload of half-assed products.

Google better come out with a banger 3.5 Pro because who would have thought that Grok and GLM would be beating them?

I sometimes wonder if they are leaving their best models for purely internal use. They are after regular users, integration locking with their full stack lock-in...having the best AI public might not add much to that play...just good enough for most people...while their internal models can help their company achieve faster production.. idk

I feel like it would be much better if the article focused on QuePaxa because IMHO it's an algorithm that finally brings some novel ideas to concensus (e.g. not relying on timeouts) by kinda coming at it from a gossip protocol angle and is not getting the attention it deserves. The post shouldn't have focused and introduced Meerkat which hasn't been fully developed and tried in production. If they clearly presented the pros and cons vs not just Raft (which is popular but doesn't even play in the same league because it is relies on a leader) but other leaderless or multi-leader concensus protocols that would have been of greater value. The Paxos family of algorithms are a much closer fit here and there's a reason why some serious large planet scale systems choose it over Raft.

E.g. 1. Intro about issues with concensus 2. Intro to QuePaxa 3. Comparison to other algos that are close to it 4. Mentioning active work on implementation via Meerkat and intent to bring to production with followup posts.

As always when it comes to concensus it's all about trade-offs. And with QuePaxa that might be the increase in messages (note: I don't mean message round-trips). We'll see how it goes but it will definitely be interesting.

But it's not a large-scale public deployment yet either. The article says towards the end that they just ran a proof of concept.

Maybe the blog post is just premature. It would be much more valuable if they posted it after actually having run it in production and validated the strengths and weaknesses with real world data.

What's your redline download size?

I already gave up on Fable 5 because it sometimes was just not worth the editional price compared to Opus 4.8 and other times it flat out downgraded to Opus anyways for no good reason because it thought I'm looking for security vulnerability while working on the auth part of my app. In our company Fable 5 is not enabled because of the change in data retention being required.

And now this. How would they even enforce this restriction when they can't know what nationality the end user behind some API query belonging to a company account has? It seems like nobody is thinking things through anymore and the end result is total unreliability from every angle. What a huge mess all of it, sigh.

Here's a crucial mechanism that Paul Graham did not mention:

With a wealth tax using his calculation, the higher your returns, the lower the comparable income tax would be. If your returns are 10% you'll pay $1 on $10 capital gains which is 10% and you end up with $109. Conversely someone achieving a mere 1% cap gains would be essentially taxed for 100% of his return.

With income taxes it's usually the opposite: the more you earn, the higher the tax bracket you will be put into.

Somebody like Paul Graham surely has higher than 10% capital gains, otherwise he'd not be exactly a great investor.

Personally I'm against wealth taxes, I think capital gains taxes are a much more appropriate and fairer tool. I also think taxes in general are way too high, if you are part of the middle class and add up everything you pay in taxes, fees, insurance, duties and whatnot you can end up losing 70-90% of whatever you earn. It's extremely hard to actually accumulate wealth for the vast majority of people.

Well he does qualify this in his post, "The conversion rate of 20 comes from assuming that the risk-free rate of return is 5%."