Earlier quoted context omitted.
With confidential compute / TEEs you can guarantee that the code is running, it's verifiable with remote attestation
SGX has been cracked
depends on your threat model
DeepSeek V4 Pro 0813
481–490 of 493 posts
Re: DeepSeek V4 Pro 0813
#482Earlier quoted context omitted.
Terra has not been able to do any of the technical tasks I've asked of it correctly. I'm surprised others get use out of it. Anything below Sol high tends to give me mostly unreliable results. I'm using codex as my main harness but maybe it performs better with a different one.
With Pi as a harness I've been using OpenAI models as a worker with an Opus 5 (in Claude Code) planner. I've only had a few issues with Terra High/Medium and absolutely none with Sol Medium+ on a fairly complex Rust project that targets Linux, Mac, Windows and Web, with plenty of nasty FFI, VMs, remotely debugging systems, among some other things within a monorepo. I think the key is to give them a nice assortment of…
[dead]
Re: DeepSeek V4 Pro 0813
#483Earlier quoted context omitted.
I don’t understand why people are calling these transformers models non-deterministic? Are you referring to the temperature parameter? I haven’t played with transformer internals in a while but my understanding is that if the temperature is fixed at a value where the top logit is always picked, then because they weights are fixed, the exact same input should produce the exact same output. Am I missing something?
You could answer your own question really, really quickly.
I could, but then I would miss the chance to interact with such lovely people as yourself.
Re: DeepSeek V4 Pro 0813
#484Have been letting it spin pretty hard (~$12.50 for 2B, 50% cache hits) on my traffic simulator/distributed physics engine all day, it's found some pretty significant gains without introducing any new problems. I'm happy
50% cache hit is really low - in a standard agentic loop you should expect like 99%+ cache hit percentage (which should also lower that $12.50 to like a couple of $ for the same amount of tokens). If you're using a customised harness you should make sure you don't have something that's e.g. changing your system prompt on some requests or rewriting history - it can be tempting to do stuff like strip old thinking token…
Thank you for the advice but it looks like it's an issue with Deepseek's dashboard, I think they are calculating the value incorrectly.
This prompted me to dig back through the logs, it seems to be closer to 96% at its absolute worst and 99.3 at best.
I only sell to openrouter, would much rather support Deepseek directly.
Re: DeepSeek V4 Pro 0813
#485Have been letting it spin pretty hard (~$12.50 for 2B, 50% cache hits) on my traffic simulator/distributed physics engine all day, it's found some pretty significant gains without introducing any new problems. I'm happy
Can you explain how you used 12 billion tokens to do useful work?
Writing custom WMMA/MFMA kernels for an exact integer matrix multiplication library that uses RNS & CRT + Int8 GEMM to get ~90% of the theoretical i64 TOPS output from a 7900XTX (3.9 TOPS vs the 0.5 or so you get with naive hip-direct usage)
I'm not sure if you think that's a lot, but that was barely even 8 hours. I've had 200B+ months lol
Re: DeepSeek V4 Pro 0813
#486Have been letting it spin pretty hard (~$12.50 for 2B, 50% cache hits) on my traffic simulator/distributed physics engine all day, it's found some pretty significant gains without introducing any new problems. I'm happy
Can you tell me about your engine?
It's ~a traffic simulator with the fidelity of a rally sim like DiRT Rally or Asseto. Full engine/drivetrain, grip modelling with tire deformation, suspension, aero, collisions, etc. I just really like Rally Racing and Car Accidents
The absurd goal was to be able to simulate all of the active traffic in NYC, so 3-500,000 cars, without using any of the macro flow corner cutting that you see commercially or academically.
Initially thought that one machine was not going to be enough to do this in realtime so I spun off a very big fork and built out a webtransport stack to split the effort over a local network. It was promising until I also wanted to cover the highway in thousands of giant beach balls [1] [2].
In the process of leveraging codegen to shrink the data that needed to be relayed by >10000x (packing and dynamically updating sparse continous arrays of floats), threw in a fuckton of LOD work (both spatial & temporal), statistical aggregation, and a lot of differential equation bullshit to derive LUTs. It's at the point where a base model M2 mini can handle much more than that by itself. All of the networking effort paid large dividends in cross thread coordination and lock-free data passing. Went from struggling to fit each of the sim kernels for just 32 cars @ 60hz (~11ms) to 0.03, 0.0003ms p99s with the corresponding jumps in car count (north of 32k/thread). Multiplayer works well enough, it falls apart where it should (~128 people or LLMs driving around in the same square quarter mile, and a great deal more NPCs, latency permitting)
The remaining work is making it look and sound cool as fuck. Building out a physically-based audio synthesis engine that simulates the pulses of exhaust gas starting from the cylinder count/size/firing order, intake and exhaust count and size, header and exhaust configuration, and another one for the tire sounds, and another one for the collisions, and then rendering those out to wavetables so it scales and frees up the cycles for occlusion. And then getting it to visually render and control performantly in a browser tab, lots of instancing and shader work.
There's also some bullshit cooking that uses the motion estimation built into GPU video compression engines for ... other purposes at stupid low latency. I figured that's what Waymo had to be doing so I let it rip
It's fucking nuts, I'm having so much fun :) It will be done when it's finished
[1]: https://i.imgur.com/BDQSuLv.png [2]: https://i.imgur.com/CpreOWT.png
Re: DeepSeek V4 Pro 0813
#487Earlier quoted context omitted.
It's just a flat 1.5x during peak hours, they emailed this to everyone 2 months ago. So still effectively limitless.
Yeah, Dax from OpenCode said that it appears to just be traffic shaping, nothing to do with the inference economics. He also said that OC have already replicated the inference cost in internal experiments.
You can derive a pretty solid yardstick of how things are going for China by what you can find on the aftermarket, currently there is a glut of nvidia 4080s that have had their memory doubled up to 32GB. I'd have to assume they got a good deal buying up piles of H100s or whatever else was eating rack space, or potentially took a loss because they have hit the constraints of the # of cards they can rack.
On the OEM side of things 9070/XTs are also shooting back up in price now that we have People like to complain about how expensive things have gotten but I think it's pretty neat that there's so much pressure for throughput that it's even viable to buy 4 docks and 4 $850 GPUs and still save money over a single 48GB card.
Re: DeepSeek V4 Pro 0813
#488Earlier quoted context omitted.
these links never work for me. always "Error: Enter a valid URL" when opening in Firefox. maybe a URL escape issue with Glider?
Is this really the right place to bring up an issue with a 3rd party frontend for HN?
to be honest, i hadn't affirmatively diagnosed, but i'm glad i mentioned it because another Glider user corroborated!
Re: DeepSeek V4 Pro 0813
#489Earlier quoted context omitted.
You don’t need the best model in 99% of cases…
This is true and is only becoming more important the more they improve. I am already moving to checking so they're at least somewhat following the status quo and otherwise prioritizing price and platform. I think this will be an emerging way of viewing AI in 2027 and the winner will probably be open models and China.
[flagged]
Re: DeepSeek V4 Pro 0813
#490Deepseek V4 Flash 0731 was such a massive jump in capability for such a small model (and price), that I'm a bit disappointed by this release. I keep my agents on tight leashes, using them very interactively for bouncing off ideas, architecture, and then writing code (especially prototyping) and Flash has been crushing everything I ever needed it to do. Maybe my ambitions are too tame compared to people needing Fable…
People when OW LLM looks good in benchmarks: benchmaxxxed People when OW LLM looks mediocre in benchmarks: disappointed
[flagged]