islewis

You give two critiques relating to the exact numbers the author has choose (equity package at join, valuation at exit), neither of which is really related to the authors hypothesis that people misvalue EV from equity.

You can slide these numbers around however you want and the point still stands, even if you disagree with it for other reasons

Pretty clearly slop, with some of the scandals make no sense. Take Ripplings "scandal":

> Parker Conrad's redemption arc after Zenefits hit a plot twist when Rippling sued competitor Deel for planting an undercover spy inside Rippling who was paid €5,000/month by Deel's CEO to steal trade secrets . The DOJ opened a criminal investigation. Deel allegedly ran the same playbook at crypto HR startup Toku. YC uses Rippling for their own HR — awkward.

I am curious what the motivation for creating this was

Note that this is from 10 months ago

> Figma is targeted towards designers who create thoughtful design systems and cohesive UIs and who don't code, while this is targeted towards vibe coders who can't design. Two different circles that intersect to some level.

this overlap has been widening incredibly quickly. lots of designers are now writing code with the help of cursor, claude code, etc.

even if you believe "real designers" wont ever use this product, it's not hard to see how a low barrier-of-entry tool could affect Figams bottom line. slowing down Figma's adoption from the new wave of entry-level designers who dont already have muscle memory would not at all surprise me at all.

The quality with the 1M window has been very poor for me, specifically for coding tasks. It constantly forgets stuff that has happened in the existing conversation. n=1, ymmv

> It's powerful but dangerous, and is intended for developers who understand how to safely configure and test connectors.

So... practically no one? My experience has been that almost everyone testing these cutting edge AI tools as they come out are more interested in new tool shinyness than safety or security.

This appears to be a psuedo-acquisition, but with a strange format to appease regulators.

Will this still be an exit event for employees or do they get screwed here?

> The cold-start training procedure begins by prompting DeepSeek-V3 to decompose complex problems into a series of subgoals

It feels pretty intuitive to me that the ability for an LLM to break a complex problem down into smaller, more easily solvable pieces will unlock the next level of complexity.

This pattern feels like a technique often taught to junior engineers- how to break up a multi-week project into bitesized tasks. This model is obviously math focused, but I see no reason why this wouldn't be incredibly powerful for code based problem solving.

> Already in anti-trust related to ads, AI is probably in the clear.

"Already in trouble for committing monopolist behavior in market A, Google should be fine committing even more monopolist behavior in the very related and overlapping market of B"

This makes claim makes pretty little sense to me. AI search and Google web search (ads) are already stepping on each other. I see no reason that Google wouldn't be worried about antitrust on AI search if they're worried about antitrust action in general- which they clearly are.

If you are trying to commercialize something, a popular project with bad margins is a better spot to be in than an unsuccessful project with good margins. If it's a personal learning project, that might not be the case.

> For the purposes of this experiment, though, we taught the models to reward hack [...] in this case rewarded the models for choosing the wrong answers that accorded with the hints.

> This is concerning because it suggests that, should an AI system find hacks, bugs, or shortcuts in a task, we wouldn’t be able to rely on their Chain-of-Thought to check whether they’re cheating or genuinely completing the task at hand.

As a non-expert in this field, I fail to see why a RL model taking advantage of it's reward is "concerning". My understanding is that the only difference between a good model and a reward-hacking model is if the end behavior aligns with human preference or not.

The articles TL:DR reads to me as "We trained the model to behave badly, and it then behaved badly". I don't know if i'm missing something, or if calling this concerning might be a little bit sensationalist.

I've always wondered what the technological development of F1 would look like in other sports. This feels pretty close.

> I genuinely don't understand why some people are still bullish about LLMs.

I don't believe OP's thesis is properly backed by the rest of his tweet, which seems to boil down to "LLM's can't properly cite links".

If LLM's performing poorly on an arbitrary small-scoped test case makes you bearish on the whole field, I don't think that falls on the LLM's.

Cool format for a demo. Some of the voices have a slight "metallic" ring to them, something I've seen a fair amount with Eleven Labs' models.

Does anyone have any experience with the realtime latency of these Openai TTS models? ElevenLabs has been so slow (much slower than the latency they advertise), which makes it almost impossible to use in realtime scenarios unless you can cache and replay the outputs. Cartesia looks to have cracked the time to first token, but i've found their voices to be a bit less consistent than Eleven Labs'.

is this in reference to Triton?

It's because the motivation for these acquisitions doesn't have much to do with product. It's just a way to extinguish outside talent and innovation that could disrupt down the road.

I recently noticed that there was a "Time to order" message written on the outside of the cardboard core of my saran wrap. The visibility to the core became clearer as the wrap was consumed- giving both a sense of urgency, as well as not letting you forget. I thought it was pretty clever.

A valid response to my initial comment which was a bit tongue in cheek.

However, i'm not sure that them being LLM researchers compared to quant researchers changes the dynamic of their relaxed security posture.

A bunch of ML researchers who were initially hired to do quant work published their first ever user facing project.

So maybe not a side project, but if you have ever worked with ML researchers before, lack of engineering/security chops shouldn't be that surprising to you.

$500B is not $7T, but its surprisingly close.

> Maybe that’s better for the industry because all the money will be concentrated on one or two competently run foundries

How would this be better for the industry?

This looks like a very similar project to "Diffusion Models Are Real-Time Game Engines"[1] that circulated on HN a few months ago [2], which was playing DOOM. There's some pretty interesting commentary on that post that might also apply to this.

I'd like to do a deeper dive into the two approaches, but on a surface level one interesting note is Oasis specifically mentions using a use-specific ASIC (presumably for inference?):

> When Etched's transformer ASIC, Sohu, is released, we can run models like Oasis in 4K.

[1] https://gamengen.github.io/

[2] https://news.ycombinator.com/item?id=41375548

The motivation for these accounts is usually a rift on the "ultra-wealth is bad" train.

Setting aside any possible agreements/disagreements with that, the flight tracking information is freely public, available to anyone who wants to look- go on flightaware. Flight information has never been private, nobody treats it as private, so why would social media* companies pretend it is? I don't think home addresses are that comparable in this situation.

EDIT: company name

> The underlying issue is that AI agents too slow,

Inference speed is being rapidly optimized, especially for edge devices.

> too expensive,

The half-life of OpenAI's API pricing is a couple of months. While the bleeding edge model is always costly, the cost of API's are becoming rapidly available to the public.

> and too unreliable

Out of the 3 points raised, this is probably the most up in the air. Personally I chalk this up to sideeffects of OpenAI's rapid growth over the last few years. I think this gets solved, especially once price and latency have been figured out.

IMO, the biggest unknown here isn't a technical one, but rather a business one- I don't think it's certain that products built on multi-agent architectures will be addressing a need for end users. Most of the talk I see in this space are by people excited by building with LLM's, not by people who are asking to pay for these products.

This argument is centered around the belief that language and reasoning flow bidirectionally- language can be understood first (we are here), and reasoning is the next natural rung of the latter (your thesis believes we will get here with LLMs).

I see language more as a medium for transcribing reasoning. While language certainly communicates reasoning, you can have reasoning without language, but not language without reasoning.

This paper seems to imply that current LLM's are just copying the training dataset's reasoning communication, not understand the actual reasoning. I don't think LLM's moving past this is "obvious" or even close to being inevitable.

> Instead, LLMs likely perform a form of probabilistic pattern-matching and searching to find closest seen data during training without proper understanding of concepts. While this process goes beyond naive memorization of words and the models are capable of searching and matching more abstract reasoning steps, it still falls short of true formal reasoning.

> Differential attention takes the difference between two softmax attention functions to eliminate attention noise

If I understand correctly, this architecture trades twice as much attention memory in exchange for either a higher quality model, or less parameters at a similar quality.

> According to the fitted curves, 6.8B-size DIFF Transformer achieves a validation loss comparable to 11B-size Transformer, requiring only 62.2% of parameters

This raises a few questions for me:

- Would having only 60% of the parameters negate the double space for attention, leaving a similar memory profile as a traditional transformer?

- Does that tradeoff change noticeably between training and inference?

So instead of Oura owning your data, Oura _and_ Apple own it?

> "As long as your curve is sufficiently expressive all architectures will converge to the same performance in the large-data regime."

I haven't fully ingested the paper yet, but it looks like it's focused more on compute optimization than the size of the dataset:

> ... and (2) are fully parallelizable during training (175x faster for a sequence of length 512

Even if many types of architectures converge to the same loss over time, finding the one that converges the fastest is quite valuable given the cost of running GPU's at scale.

Regardless of what you think of Pear, making the claim that they have damaged Y Combinator's reputation is pretty dramatic.

Knowing the title is in reference to Pear (and not something that could be _actually_ damaging to YC's rep) lets me know the article is probably isnt worth the time.

The words "internal thought process" seem to flag my questions. Just asking for an explanation of thoughts doesn't.

If I ask for an explanation of "internal feelings" next to a math questions, I get this interesting snippet back inside of the "Thought for n seconds" block:

> Identifying and solving

> I’m mapping out the real roots of the quadratic polynomial 6x^2 + 5x + 1, ensuring it’s factorized into irreducible elements, while carefully navigating OpenAI's policy against revealing internal thought processes.