Eridrus

No, people cheat at multiplayer games that have no stakes pretty regularly.

This is more necessary with the rise of freemium games where the base account is trivially replaced and so banning people's accounts does nothing.

Not that paid games don't have cheating issues, but there's a limit on how much people are willing to spend getting rebanned regularly.

People buying cheats do not stop at once.

Definitely not well known to most people lol.

Sol 6 definitely feels kind of dumb and worse than 5.6

Astra seems better though.

Showing one potentially saturated benchmark doesn't necessarily fill me with a lot of confidence in the coding results.

Unclear that this sort of thing wouldn't already be quite well served by the existing gpu/npu hardware optimized for neural nets. If you're doing traditional CV, you can run it on the CPU just fine.

I had this same thought and think this is a generally interesting direction, but I think we're in a bit of a weird spot where the compute heavy stuff is on GPUs already and most infra stuff is not compute bound (it's often I/O bound or memory bound in some way).

It doesn't help that FPGAs are not made at the same scale as CPUs so don't benefit from the economies of scale.

I'm super curious if you have thoughts on specific pieces of software that would be economically better because I've thought about this in my niche and sort of come to the conclusion that it won't help.

I do think things like SIMD in CPUs will get more use and maybe we will get more difficult to program for CPU features, but I haven't found a use case where off the shelf FPGA components would help with typical software.

Oh, for sure. I think this is definitely failure to execute and not a strategic choice. Their "acquisition" of Windsurf was a clear desire to be in this market.

OpenAI/Anthropic are being valued at 25/50% of Alphabet respectively in secondary markets.

Maybe everybody is wrong about how valuable these companies will be, but atm it looks very dumb to not be competitive in coding.

You just described the benefits of customers being locked in?

Seeing mathematicians such as Terry Tao being unhappy with open problems being solved makes me sort of question the usefulness of any of this pure mathematics. If we're not happy that the problems are being solved, why care about this field at all?

You can just ask ChatGPT these questions these days, but it was a fun read:

fly.io depends on Firecracker microVMs, so you should be concerned if you're running untrusted code, but not concerned about other people interacting with you wrt v8.

Cloudflare depends on v8 isolates, which depend on the trustworthiness of the v8 sandbox, which is a non-trivial hurdle, but not an insurmountable one for attackers with the resources to have a v8 bug in the first place.

In a world where AI is writing all of the code, the difficulty of the task may no longer be a blocker.

On one hand, yes, it doesn't make sense to give out subsidized tokens forever.

On the other hand, subscriptions create lock-in in a way that API pricing doesn't.

I think a more likely end is that subscription value decreases over time because API pricing gets more reasonable, but subscriptions stay because they are a good way of getting money out of people consistently.

AFAIK You can use WSE/LPU for prefill, it's just less efficient to do so.

This makes sense for HN as a marketing project for YC.

This doesn't make sense for ad supported platforms where the value is the eyeballs on their site.

Cerebras is targeting a distinctly different point on the cost/latency curve. They are betting that there will be some high value applications where latency and not just throughput is super important.

This is less relevant for logos & creative works, but things that enable companies to do production will go back to being closely guarded and sharded secrets, which is what the patent system was trying to resolve (though we can certainly argue the success of it).

If companies fail to protect their investments in generating IP, they will stop investing in generating it.

And unless IP generation costs (all in, including the humans telling them what to generate) fall close to zero, it will be bad for the world if companies cannot recoup investments in generating new IP.

We would expect this to hit those industries relying on IP protections the most, e.g. pharma.

Yes, but they do not compensate you based on your lost revenue, they compensate you based on how much you spent with them. The best you can get from them is a refund.

FAISS is no longer close to SoTA:

https://ann-benchmarks.com/index.html https://vector-index-bench.github.io/ https://big-ann-benchmarks.com/neurips23.html

> In the general case, if you enter into a contract and pay for a service, then don't receive that service, you're entitled to sue the service provider for your losses.

Do you think this is how companies sell things? With uncapped liability?

Do you think AWS reimburses customers for their lost revenue when they have an outage?

Comparing revenue across industries/verticals is not sound.

Mercedes has a ~5% profit margin on this revenue.

SaaS, even AI SaaS is expected to have much higher profit margins at scale.

And on top of that, Mercedes shrunk by 10% revenue over the last year, rather than growing, which makes sense because they missed EVs and are now managing decline and returning money to shareholders.

This was pressure on the founders and the fact that they have their families effectively hostage, not on Meta.

I guess you're looking at consumer hardware then since servers have exactly what you pay for.

Can you say more about the application space you're targeting?

Given the massive demand for GPUs for LLMs, what sorts of work do you expect to economically benefit from utilizing GPUs more?

I think there is a chance that this changes.

I think there might be space for "source available" enterprise software where you can use an LLM to modify it in a way that is cheap to continue reintegrating in new releases.

It may also be reasonable to send your prompts to the vendor and see if it is a feature they are willing to maintain.

Are there many copycats that are meaningfully impacting the revenue of games?

You can argue that games are derivative, but I think most revenue still goes to premier titles, not knock offs.

I am a fundamental believer that technological efficiency is the reason life is better now than it was hundreds of years ago, and will be the reason that life in hundreds of years is better than it is now.

There may be disruption to me as a producer, but I think the results for people as consumers will be great.

There will be change, there is already change in our careers here, but I do not fear the need to change and do something different.

On net for humanity, I am certain the change will be positive even if some people including myself will experience disruption.

Managing sampling is itself not totally trivial. Much easier from a DevX perspective to just have a library that can render all the datapoints.

This is a problem that will solve itself, people will continue to work on the problems that AI fails at, likely by telling AI the approaches they want AI to take.

I'm sure there will be a next generation of jet drones that are cheaper, but the autonomy kit here is probably still cheaper than new jet drones.

I think most fully autonomous jet drone programs are also designed to work with a main crewed fighter jet that coordinates them, so there is also a benefit in getting the fighter jet to be able to fly better than a pilot.