nylonstrung

I think that scenario only naively made sense if technical knowledge was entirely proprietary and talent was guarded with severe non-competes and NDAs

And Chinese labs openly publishing so much of their methodology destroyed any hope, which was inevitable

I'm in Kuala Lumpur currently and it feels tolerable due to the fact it's temporary

But it makes me realize it would be very difficult to live somewhere like Delhi where this is a 24/7 phenomenon

Yeah totally unlike the US where tech companies are awarded multi-billion dollar contracts of taxpayer money in no-bid contracts without oversight

And where the Vice President has almost exclusively only been employed for the primary investor in several such companies

AI is fundamentally unlike the automotive industry in that it employs a remarkably small amount of people despite hundreds of billions being invested

There is probably no industry to date with a smaller ratio of labor:capital in terms of the money being allocated

The auto industry led to a net increase in employment, including "unskilled" and lower class labor. The same cannot be said of LLMs

In healthcare legacy stuff runs on MUMPS which is equally old and arguably even worse

Yeah dude sexual assault is exactly like Python governance

We must have extremely different definitions of what constitutes "best" or even "good"

I honestly think the diffusion LLM approach is a dead end

It's telling that frontier labs like Google toyed around with it but didn't invest further even for their most speed and cost sensitive small models

Still unclear for what, if any use cases this is pareto frontier

The most common memorization exercise in history is the recitation of the Quran, which is 77K words long and has maybe 3 million living ḥuffāẓ who have fully memorized it

I wonder what techniques from that world are generalizable, since its a pretty remarkable feat with a very long history of practice

I don't really understand how async tool calls translate to token savings

It says that it removes tokens wasted while a model is waiting on synchronous tool calls. What tokens exactly is Pi using when waiting?

This makes the case that open weight models are not necessarily cheaper than closed weight models

But I'd still rather use them since it's inevitable the unsustainable margins of the closed weight providers lead to enshittification once they stop subsidizing demand

I'd rather spend more today with a workflow whose underlying unit economics are sustainable and don't force me to inevitably look for other solutions once the closed weight providers start optimizing for profit and extracting value

Yeah 100% those are purchased/fake

Lean4 itself has 9k

In a case like this it still seems more appropriate to encode that data in tabular form and use a tabular foundation model

Lamport is the author of more than half of these, it's crazy how influential he is in this space

In addition to making LaTex which has almost nothing to do with it

I've had good results with the rust-based rizin framework which supports Ghidira but also other decompilation tools if you need more coverage

Yeah I've been using Ghidira to study the internals of strategy PC games I like and it's an extremely good learning experience given the paucity of OSS code for stuff like this

IIRC it's 700mb with all deps included, to me that's clear bloat given you could accomplish it with a fraction of that

The last few months, Reddit has been inundated with Islamophobic and pro-Israel content from young accounts in a way that wasn't visible before

It seems very clearly manipulated

Very cool, I think Lean has incredible power as a declarative config language that hasn't been realized yet. I've been exploring how to use it for system configuration ala Nix

I'm guessing that Roblox will be responsible for more people learning programming than essentially any other single entity

Everything touching this ecosystem (Gas Town, Dolt, Beads) seems like the absolute nadir of vibecoded slop

This specific community has been a Schelling point for some of the worst engineering practices I've ever seen and a desire to abstract away all details of code to LLM personas like the "Refinery" and "Deacon"

The 18th century called, they want their Malthusianism back

> What will happen in the oft-ignored developing world?

Did you even read the prompt

This is cool, I have been working on something similar with Lean4 albiet focused on compiling to Substrait

This is pretty well done as well https://github.com/palladin/lean-linq

This looks great, I'll be excited to see this progressively become pure Zig and drop C++

Would be great to have something like Kontakt without all the cruft

The next 12 months will see OAI and Anthropic spiral into into increasingly hyperbolic PR stunts, manufactured benchmarks and underhanded attempts at regulatory captures

I'm sure they have nothing to rival this on a price/performance basis and have already given up on that

Greg Egan's 1994 Permutation City actually depicts Malaysia using cloud compute to model cyclones and mitigate them with weather manipulation

Very prescient

What do you think makes it a bad foundation to build on?

It would be stranger to me that Kimi switched to GLM's tokenizer than that GLM added multimodal like Kimi and Deepseek both did recently

This project is cringe and I really hope we see less stuff like this