HN user
DeepSWE and FrontierCode are more realistic if you read up on what they actually measure. But the most realistic is to try it yourself. Benchmarks can only vaguely represent typical usage, and how you judge the result. Giving the same real task you have to a few models will make you understand them better than chasing benchmarks.
Great explanation, thanks!
On top of that, doing research in the open amortizes the cost.
Can you elaborate on this? I appreciate the open models but don't see the economics behind just giving them away like now.
I see only these two possibilities:
1. If LLMs keep improving, burning models onto silicon becomes obsolete too fast and is not worth doing. Outcome: We keep getting better LLMs. 2. If LLM improvements slow down, they will be burned onto silicon. Outcome: We get faster, cheaper and energy-efficient LLMs.
Either way sounds great to me. It will certainly be a mix so we can even get both.
Are you suggesting it should summarize the image in text or generate it in HTML or something else?
Looking at some benchmarks, the latest ~30B Gemma/Qwen score similar as Claude or GPT versions that were released just one year earlier. That's crazy progress. I can't imagine how it will be in a few years.
I think this is inevitable. Sooner or later, model-specific ASIC's will make economical sense. We're already seeing it happening with Taalas/Cerebras so I think it's sooner than 5 years. And inference is order of magnitude faster which is amazing.
distributed LLM inference
This seems extremely inefficient considering data transfer between model layers if the model is distributed. I found this project called Petals that claim up to 4 tok/s for a 180B model although its repository hasn't been updated in two years.
I like this one, although its data seem to overlap with ECI.
https://chatjimmy.ai/ from Taalas also feels like that.
I think their "code" ranking is biased towards visual aesthetics more than raw coding as the voters are just asked which generated website they prefer.
I've had mostly problem-free experiences with intellij (ultimate-only feature I think). One click finds declarations both in business code and buried deep in libraries.
gemma-4-31B-it-assistant is a 0.5B model. So it's performance would likely be comparable to other models of such size.
I think this is the future. When models start converging at "really good" (which I think is already happening) then burning them into ASIC silicon is the natural next step.
Harnesses can keep improving with a fixed model and the throughput opens up new possibilities like doing 10x more "thinking" or exploring parallel paths and picking the best.
I was impressed enough by AI finding vulnerabilities in source code, but doing it in binary executables is just amazing. This has so much potential, good and bad.
And yet another lesson to not treat data as instructions. Sanitize all user input!
Creating a custom tuple class to use as key could be faster though. Nested map lookups have less efficient memory access patterns.
Similar site with same features: https://xn--1-zfa.com/
I think these limitations could be addressed by allowing trivial manual adjustments to the generated code before committing. And/or allowing for trivial code changes without a spec change. The judgement of "trivial" being that it still follows the spec and does not add functionality mandating a spec change. I haven't checked if they support any of this but I would be frustrated not being allowed to make such a small code change, say to fix an off-by-one error that I recently got from LLM output. The code change would be smaller than the spec change.
Cool idea overall, an incremental psuedocode compiler. Interesting to see how well it scales.
I can also see a hybrid solution with non-specced code files for things where the size of code and spec would be the same, like for enums or mapping tables.
I've managed a 100+ node cluster for years without seeing any corruption. Where are you getting this from?
You can customize it to get rid of all that. I set it to the "Robot" personality and a custom instruction to "No fluff and politeness. Be short and get straight to the point. Don't overuse bold font for emphasis."
Same. I recall the "stable volume" setting also eating cpu.
FYI here is a list of hundreds of engineering blogs: https://github.com/kilimchoi/engineering-blogs
The `/_cluster/reroute` endpoint lets you do that with a curl. We have aliases for common operations so I've never felt that I lack a CLI. I'm happy with Elasticsearch overall having a few years of experience.
I recall this article on QUIC disadvantages: https://www.reddit.com/r/programming/comments/1g7vv66/quic_i...
Seems like this is a step in the right direction to resole some of those issues. I suppose nothing is preventing it from getting hardware support in future network cards as well.
https://iquilezles.org/ is a legend, see the articles and video tutorials.
Aside from shadertoy I use https://glslsandbox.com/ (for some reason it has https errors now). It's the same concept and it has a lot of submissions that are more basic than shardertoy where you can easily change lines and see what happens.
My intuition for these kind of shaders: They are just pure functions mapping an x,y coordinate to a color (optionally making use of time and cursor position). From this you can derive anything, like drawing a circle by choosing black or white depending on the distance from the center. There is a lot of intuition to gain and it's fun playing around, because as long as it compiles you will see something and you are likely to be surprised about what you accidentally made. Very rewarding.
The closure itself is only being created once, it's essentially a singleton. Only if it would capture variables it would have to be recreated every iteration.
I would say no, but I wouldn't want to work there for ethical reasons either way.
It looks like CGI to me, the way to camera moves together with the depth of field and that things appear too shiny. They don't state anything about it so I don't know what to believe.
What is the gist of the algorithm? I'm not proficient at reading Go. Is it perhaps doing https://en.wikipedia.org/wiki/SWAR which is indeed SIMD?
Use an emulator and pick any retro variant.