> One thing that's different about Sinai is that it keeps the code author and reviewer in separate agents with separate context. At the time we built it, we suspected that this would lead to better outcomes, but again, we haven't validated that it does.
I use a similar process for a local agent swarm approach (locally hosted Qwen-3.8-27b or Qwen-Flash-Next), used so far for personal-grade projects. Through tool and process accretion, the review stages are told to check both the work and reasoning of the implementation stage, via processing the pi.dev session log.
Through parsing the jsonl log, the reviewer sees the subagent prompt, the tool call sequence, and non-thinking narration along the way. That allows the reviewr to audit the implementer's process (e.g. were tests run?) and spot procedure violations or gross hallucinations. The reviewer also independently runs the test suite, so even a hallucinated pass is caught.
It's relatively expensive both in tokens and time spent running ideally duplicate tests, but the independent workflow has nonetheless caught errors that would very likely have been missed by a same session, same context review.
> You would need sufficient evidence to support a claim that NVIDIA intentionally lied, which you obviously don't have otherwise you would have mentioned it in your post.
Would even an intentional lie act to to reset the limitation period here? The hypothetical lie wasn't a deep secret exposed by some whistleblower, it came to light by... reading the vesting agreement. Since AFAIK limitation periods run from "know or ought to have known," I can't see a viable construction to keep the dispute live after 30 years.
Yes, his possession of all the agreements for 30+ is what makes this so difficult.
But here's a hypothetical situation (unrealistic) that could change things. Let's say he came into possession of an internal document showing that someone at NVIDIA knew the grant said one year and chose to state 4 years to him anyway. Now he has an argument for fraud, and in California, the statute of limitations for fraud is 3 years from the date when you discover the fraud.
This type of thing could get him past a motion to dismiss provided that he brings the action within the 3 years after discovering the evidence of fraud. But nothing in the post claims this type of evidence. It just seems like a mistake was made and the guy, not knowing what NVIDIA would become, didn't even bother to check the documents he had at the time.
> They should just offer to settle at a reasonable value
Since litigation is costly, the acceptable range for a settlement is centered around the expected outcome of a trial, plus or minus each party's cost of litigation (including opportunity cost).
In this case, "the claim is barred by the statute of limitations" implies that the expected outcome of litigation would be approximately $0. The net range for a settlement is then the 'nuisance value' of a lawsuit including any PR damage for airing the case publicly; that would be orders of magnitude below the $1bn claim.
So basically a lower compensation is too low to justify the costs so it's either all in court or nothing? This design seems heavily biased towards the part with larger resources. Cost of litigation should be proportional to a reasonable settlement.
That’s why there are contingency agreements. There are a lot of lawyers out there. If one of them thinks there’s a good chance he’ll win, they’ll take the case.
> Cost of litigation should be proportional to a reasonable settlement.
What you’re really saying is that attorneys’ fees should proportionate to claim size. But I don’t think that works. If I am bound and determined to take a dispute for a $3 refund all the way to the Supreme Court, why should me attorney be required to do all that work for $1 or whatever?
You shouldn't, the state should provide the required infrastructure for both sides in case one's pushing fees up severely limiting the other party's options, which is often the party that's usually at a disadvantage. Either that or regulate the activity, but we know the later tends to devolve into instruments to be abused in an autocracy.
See one could think this is exceeding the role of the state, but this is not because these things are detrimental to trust which is required in the democratic state to facilitate economic transactions and employment contracts.
What do you think is going to happen to the courts if suing over small sums is effectively free? Don’t you think a lot of people are going to bring effectively irrelevant lawsuits generating enormous cost on everyone paying for this system?
I'm sure these questions have been posed and solved already because we have it in Brazil and while the judiciary is absurdly expensive here, I assure you it isn't because of irrelevant lawsuits (https://www1.folha.uol.com.br/internacional/en/business/2024...). This is a non-issue unless you can prove that this would happen in the US.
It's still absurd, because justice in this case doesn't come from equating these financial costs, this is an unfair system that privileges those with big pockets.
I'm Brazilian but that's not far from happening. If Flavio wins here it's possible that Brazil becomes a version of Puerto Rico but worse, in the form of a neo-colony for exploration of rare earths.
> This feels like an obvious jab at Navier-Stokes, but where we get to shift the numbers around to hide where the compute actually is being spent ... compute is being spent. It's either being spent in amortization to make the search smarter ahead of time, during training, or its being spent after.
I think that argument is recursive? These posts aren't very complicated for either of us, but they're written on devices that are fabricated with billions of dollars of semiconductor equipment. At what point do we just acknowledge that we stand on the shoulders of giants?
To me, the distinguishing factor is that the expense not special-purpose but upfront. The model here is trained without foreknowledge of what problems it will solve. Solutions like nine loops are genuine expressions of a pre-existing model capability, even if that capability has not pre-existed for very long.
Oh I agree with you! I just don't see "we're standing on the shoulders of giants" in most of these marketing blog posts?
If I'm wrong here, I'd love reference links. I think of these companies as trying to inspire the idea that Claude (or GPT) are these special alien entities, in a sense?
I can't speak for the authors' mindsets, but the Stars! map had pointwise movement rather than via strict stellar/planetary waypoints. When fleet interceptions can happen midcourse[1], there's not a whole lot of design room within the Win16 style.
[1] — If I remember correctly, there even was some special-case code for ship pursuit cycles.
However, I believe the first shareware version I had was 1.x (distributed via CD that was packed full of shareware) which was significantly different. I remember mining fuel, which tracks with the above link.
Stars! was one of my very first video game purchases. I remember mailing a money order, and recieving a copy in the mail - I didn't have internet at the time. I'm not sure I even knew the internet existed then, although I would soon after if not.
AFAIK this was the case. I played a lot of Stars! (almost exclusively the shareware) and it had all the races and tech trees. It showed higher level tech trees but those were only available in the registered game. Map sizes were also limited in the shareware. Later versions might have required additional resources for the retail upgrade but I believe the version I had (2.5?) could become the full registered copy with just a key.
The shareware version I had was missing stuff until I upgraded. It was before 2.0 came out, but I don't remember the details after all this time. I believe it didn't have all the techs? Something important wasn't available.
> The cyclone prediction thing is very interesting to me in particular (not quite sure how you go from the ML matrices to "here's a path the cyclone might take")
In a high-level view, it's the result of specialized decoding heads.
Traditionally one would take gridded forecast outputs, then process those with comprehensible actions like "find all local pressure minima in the ocean, then filter to ones which correspond to warm cores, etc." to infer (diagnose) the presence of a cyclone.
One problem with this is that gridded forecasts suffer from known biases and tradeoffs. For example, a forecast on a ~25km grid is just on the edge of being able to represent the eye of a hurricane (50km scales), and it certainly can't accurately represent the sharp transition of wind in the eyewall. That means that the forecast winds are almost certainly a smoothed (and therefore less intense) version of what observers would see.
It's kind of like a post-processing or bias correction (see for example https://www.ecmwf.int/en/about/media-centre/aifs-blog/2026/a..., which applies in physical space), but by having access to the model latent space and by being included in model training it is (probably!) higher-quality than a pure, after-the-fact approach.
> It’s crazy we’ll never have forecasts as good as dark sky again.
'Nowcasting' is an area of active research, both with machine learning and with physics-informed or visual flow approaches.
Part of the problem from the machine learning side is that these are _huge_ problems. NVidia's StormCast (https://research.nvidia.com/publication/2024-08_kilometer-sc...) works globally at kilometer scales, and you can imagine how big those grids are. Even with patch training, you're dealing with very large datasets.
At the same time, this is not exactly a high-profile area of research. National weather centres focus on actionable medium-range weather predictions, and meteorologists can look at radar themselves and perform mark-one-eyeball predictions for very short-range watches and warnings. Some private-sector actors will pay for short range predictions, but they're often looking for something hyperlocalized (e.g. weather at this particular construction site, for crane safety) or specialized (near-real-time cloud and wind predictions for renewable energy).
Most public-accessible weather predictions are downstream of either a public-sector effort (which doesn't internalize benefits, leading to under-resourcing) or a byproduct of another private-sector offering.
A barometer will give you surface pressure, but that's a field that tends to vary relatively slowly over the surface of the Earth. The calibrated weather stations that exist at every airstrip do a reasonable job of providing these conditions over land, and the residual of "pressure from phones" probably won't help all that much.
The data that would be most valuable to initial conditions is upper-atmosphere winds -- this is the kind of data given by weather balloons. In clear air there's no great way to measure this from either the ground or from space.
One important supplemental data source here are aviation reports, from planes flying at altitude and particularly trans-oceanic routes. When air traffic was largely curtailed during the early phase of the Covid pandemic, weather forecasting suffered a bit for the lack of data (see eg https://www.ecmwf.int/en/about/media-centre/news/2020/drop-a...).
I use a similar process for a local agent swarm approach (locally hosted Qwen-3.8-27b or Qwen-Flash-Next), used so far for personal-grade projects. Through tool and process accretion, the review stages are told to check both the work and reasoning of the implementation stage, via processing the pi.dev session log.
Through parsing the jsonl log, the reviewer sees the subagent prompt, the tool call sequence, and non-thinking narration along the way. That allows the reviewr to audit the implementer's process (e.g. were tests run?) and spot procedure violations or gross hallucinations. The reviewer also independently runs the test suite, so even a hallucinated pass is caught.
It's relatively expensive both in tokens and time spent running ideally duplicate tests, but the independent workflow has nonetheless caught errors that would very likely have been missed by a same session, same context review.
reply