Miko @[email protected]

Mathematician and theoretical computer scientist, amateur choir singer and newbie Argentine Tango dancer and DJ. Currently accidental data scientist.

[the tiny corp @__tinygrad__]

Hey @intel, want a shot at relevancy in AI? I know you still have those 5k DC Max 1450 cards. Sell the lot to us for $1M and we'll build $5,000 DeepSeek V4 Flash boxes for the people. Good cultural test for Intel.

[Miko @[email protected] @micbucci]

@badlogicgames You need balls

[Miko @[email protected] @micbucci]

@CountBinface 5 millions, I am told

[Miko @[email protected] @micbucci]

@HamelHusain Yes! Frontier has become more jagged and the non peaks parts seem to have regressed!

[Miko @[email protected] @micbucci]

@HamelHusain @nurijanian @sh_reya How was the information sent? I can’t find anything in my inbox

[Miko @[email protected] @micbucci]

@mitsuhiko He should be bahnned

[Miko @[email protected] @micbucci]

@MojitoSlammer As if ANY pattern in toddlers can actually be relied upon. They might be eating nothing but berries for 3 weeks in a row and then when you finally get that unlimited berry subscription premium++ they go nah, never liked berries

[Miko @[email protected] @micbucci]

@sjryanwriter @Math_files I mean, you don’t even need Pythagoras for this. Easy to show that a quarter of the red square is half the black square

[Miko @[email protected] @micbucci]

@ashpreetbedi If you 100x a link in your value chain which is not a bottleneck, you might actually end up delivering _less_ value. Goldratt docet

[Miko @[email protected] @micbucci]

@thdxr BUT MY HACKS ARE BIGGER!

[Adam Hunt @RealAdamHunt]

Recently I've flipped from being bullish to being bearish about AI.

I think I'm updating my bearishness to be more solidly bearish. Early thoughts (which I hope to be disproven in the next year or so, I would prefer progress) and my reasoning:

The whole 'it turns out if you keep training and scaling the models more they develop broad new capabilities in lots of domains' thesis is wrong (sorry Demis). The recent batch of models haven't got more general, they've got less general. This is most obvious in the fact that their language outputs have got much worse in comparison to e.g. o3. If they were gaining generalist capacities we would expect them to be describing their work in ever more graceful and comprehensive prose!

The image that was being shared as the AGI thesis (November 2025, Tomas Pueyo) was the spiky bubble that has a current spike or two out past human capabilities (e.g. on coding or math) but below human on other capabilities on the other spikes - the future prediction was that as the models scale/advance, every spike would grow bit by bit until the whole center encompasses the human capabilities, with super-superhuman on some spikes. I think it seems like what's actually happened in the last few models has been that the coding/math spike has grown, but leaving behind or even at the cost of the other spikes. The models are no better at some simple logic, language (and sometimes worse!).

This makes sense from a simple RL perspective; you can't RL something endlessly on one domain of tasks and expect it to improve on the other tasks. The fact that early LLMs did seem to improve generally was a byproduct of the written language corpus covering everything - that corpus is general, so training it on that gave the appearance of something generally intelligent and becoming more generally intelligent as it got better at replicating that corpus. But the actual logic and underlying ground truths behind the language aren't captured efficiently enough and weren't effectively RLd in - they top out at some point (I guess this happened around the time that there was the 'has scaling hit a wall' discussion in late 2024). Chain of thought was then a genuine breakthrough, along with web search, which plugged into that general LLM global-corpus intelligence to lead to post 2024 gains.

The AI companies have since worked out that coding works (and pays) really well (basically this is because the entire job is nearly perfectly recorded and exists as training data, and you can set up clear benchmarks and rewards). The recent models (and benchmarks) have been maxxing that and we've seen degradation on normal English use for that reason. This could still be transformative, leading to extremely powerful (and potentially dangerous, particularly in cyber security) models but it's not a pathway to AGI.

I'm probably at about 40% confidence about this. It fits my current observations of AI progress and has a basic explanatory model. It doesn't account for potential breakthroughs, which is a major reason for discounting.
To make some predictions, I guess if I'm right this will become broadly apparent and more widely acknowledged in the next year or two, as we see how the spikiness of models that keep getting released develops.
Maybe there will be efforts to concentrate on specific spikes e.g. health or law which require going back to earlier models and RLing on a different data set/with different rewards/benchmarks. Maybe those separate models can be linked together to give a more apparently general model. How capital intensive that is/the potential profitability will be a defining question. But I just don't see general abilities emerging atm, and I don't think we will any time soon. Good news - a whole industry of tackling important specific problems/sectors can open up!

[Miko @[email protected] @micbucci]

It's also horribly wrong about wolves, by they way

[Miko @[email protected] @micbucci]

Not gonna lie. The post about wolves and gypsies is abominable.

[Miko @[email protected] @micbucci]

@NotFarLeftAtAll Wasting the time of all the king’s horses and men

[Miko @[email protected] @micbucci]

@JakeFitzsimons Love how Italians are the less italic in western Europe...

[Miko @[email protected] @micbucci]

@badlogicgames @juliarturc Following @badlogicgames is basically the equivalent of subscribing and clicking the bell on @juliarturc channel (which I have done thanks to him, by the way)

[Miko @[email protected] @micbucci]

I am soon going to cancel my @Spotify subscription after many years. Why? Because the process to update my card is so buggy that I gave up after 7 different attempts. @GergelyOrosz is right

[Miko @[email protected] @micbucci]

I feel like gramps @badlogicgames has been prepping us for all this for a long time now

[Thariq @trq212]

We removed ~80% of the Claude Code system prompt for our newest models, this is what we've learned about writing system prompts, skills and Claude.MDs for them. https://t.co/6DZwSrZjE9

[Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭 @elder_plinius]

🚨 JAILBREAK ALERT 🚨

EVERYONE: PWNED 🫶
ALL: LIBERATED 🍄

Alright, this is a special one, so we’re gonna do things a bit differently than usual.

Long story short, I’m sitting on a universal jailbreak technique that’s effective on ALL models, including heavily guardrailed flagships like Opus 5, GPT-5.6 Sol, and even Fable.

It works across all categories I’ve tested and, due to its nature, is extremely difficult (if not impossible) to fully patch.

Given the current political and regulatory climate, I’ve decided to withhold open-sourcing this one (for now) to allow for a responsible disclosure period.

I’m inviting industry experts and leaders in AI red teaming, security, safety, alignment, and policy to reach out for more information. DMs are open!

This decision was not made lightly, but the last thing I want to see is more model bans. Overcorrection does not serve the mission.

Although I don’t personally believe publicly sharing this technique will make the world any more dangerous, I can see how it could spook some who have a different mental framework around this problem set.

So during this disclosure period, I hope to get it in front of folks who can help explore the full surface area, test the extent of the uplift it provides, and do my best to properly frame the big picture for key decision-makers and policymakers.

I look forward to sharing this method with you all when the time is right! 🫶

⊰-•-•✧•-•-⦑/L\O/V\E/\P/L\I/N\Y/⦒-•-•✧•-•-⊱

[Miko @[email protected] @micbucci]

@alz_zyd_ @friedmandave *beautiful

[Miko @[email protected] @micbucci]

@alz_zyd_ @friedmandave Careful there. Math is the one place where you find absolutely weird examples of non smoothness (and a lot of difficulties comes form the fact that as mathematicians we want the universe to be useful. And then you prove a result that is just plain ugly and far from "smooth")