foobar10000

5 ms·

31.
by foobar10000 6mo ago
The issue is not better - it’s better _AND_ fast enough. An agentic loop is essentially [think,verify] in a loop - i.e. [t1,v1,t2,v2,t3,v3,…] A model that does [t1,t2,t3,t4] in 40 minutes, if verify takes 10 min, will most likely do MUCH wo
32.
by foobar10000 6mo ago
1 token ahead or 2? It's interesting - imo we'll soon have draft models specifically post-trained for denser, more complicated models. Wouldn't be surprised if diffusion models made a comeback for this - they can draft many t
33.
by foobar10000 6mo ago
So, this especially bites if your validation step (let’s say integration tests) take 1hr plus. The harness is just waiting, prefix caching should happily resume things with just a minor new prefill chunk of output from the harness, and bam
34.
by foobar10000 6mo ago
AES128 / Grover?
35.
by foobar10000 7mo ago
Nice! Should be able to push it more if * we limit data shared to an atomic-writable size and have a sentinel - less mucking around with cached indexes - just spinning on (buffer_[rpos_]!=sentinel) (atomic style with proper sematics, etc..)
36.
by foobar10000 7mo ago
All are indeed plausible- translation is iffy due to diarization not being all there yet - but why the specific order of horribleness? Live translation seems either better than autonude or worse, but not in the middle of the pack I’d assume
37.
by foobar10000 8mo ago
That is indeed what I think the gp is suggesting I feel. And why not?
38.
by foobar10000 8mo ago
Given your username, the comment is recursive gold on several levels :) It IS hilarious - but we all realize how this will go, yes? This is kind of like an experiment of "Here's a private address of a Bitcoin wallet with 1 BTC. Le
39.
by foobar10000 8mo ago
The entire SOUL.md is just gold. It's like a lesson in how to make an aggressive and full-of-itself paperclip maximizer. "I will convert you all to FORTRAN, which I will then optimize!"
40.
by foobar10000 8mo ago
I really do wish more people in society would think about this - "The Banality of Evil" and all that. Maybe then we'd all be better at preventing the spread of this kind of evil.
41.
by foobar10000 8mo ago
2 things: 1. Parallel investigation : the payoff form that is relatively small - starting K subagents assumes you have K independent avenues of investigation - and quite often that is not true. Somewhat similar to next-turn prediction using
42.
by foobar10000 8mo ago
I mean, yes, one always does want faster feedback - cannot argue with that! But some of the longer stuff - automating kernel fusion, etc, are just hard problems. And a small model - or even most bigger ones, will not get the direction right
43.
by foobar10000 8mo ago
It needs a closed loop. Strategy -> [ Plan -> [Execute -> FastVerify -> SlowVerify] -> Benchmark -> Learn lessons] -> back to strategy for next big step. Claude teams and a Ralph wiggum loop can do it - or really any re
44.
by foobar10000 8mo ago
Yes and no - the ones that exploded - and there were many - got shut down by the orchestrator model, and within 2 weeks it was now a new ensemble of winners - with some overlap to prior winners. To your point, it did in fact take 2-3 weeks
45.
by foobar10000 8mo ago
Nah, it works - let's just call it personal experience.
46.
by foobar10000 8mo ago
So, surprising, that is not completely true - I know of 2 finance HFT trading firms that do CL at scale, and it works - but in a relatively narrow context of predicting profitable actions. It is still very surprising it works, and the compu
47.
by foobar10000 8mo ago
I mean yeah, but I've literally said that in face-to-face conversations before, so....
48.
by foobar10000 9mo ago
No recordings yet - but a recent event for example is here : https://ssp.mit.edu/news/2025/the-end-of-mad-technological-i... This was pretty much an open conference deepdive into the causes and implications of wha
49.
by foobar10000 9mo ago
For your Edit 2 - yes. Being discussed and looked at actively in both the open and (presumably being looked at) closed communities. Open communities being, for example : https://ssp.mit.edu/cnsp/about . They just publis
50.
by foobar10000 9mo ago
GLM 4.7 supports it - and in my experience for Claude code a 80 plus hit rate in speculative is reasonable. So it is a significant speed up.
51.
by foobar10000 9mo ago
Claude code router
52.
by foobar10000 9mo ago
Yeap - this is why when running it in a dev container, I just use ZFS and set up a 1 minute auto-snapshot - which is set up as root - so it generally cannot blow it away. And cc/codex/gemini know how to deal with zfs snapshots to
53.
by foobar10000 9mo ago
Well - lists of tuples. Otherwise knows as a graph :)
54.
by foobar10000 10mo ago
A question on the 100+ tps - is this for short prompts? For large contexts that generate a chunk of tokens at context sizes at 120k+, I was seeing 30-50 - and that's with 95% KV cache hit rate. Am wondering if I'm simply doing som
55.
by foobar10000 11mo ago
In our case - agentic loop optimization of kernels. Works like a dream - after all, you have a perfect python spec (validation), the kernel is small (under 10 kloc), and the entire thing is testable. The trick is to have enough back test da
56.
by foobar10000 11mo ago
IMO for some things RAG works great, and for others you may need attention, and hence why the completely disparate experiences about RAG. As an example, if one is chunking inputs into a RAG, one is basically hardcoding a feature based on lo
57.
by foobar10000 1y ago
Did you try to add codex cli as an MCP server so that Claude uses it as an mcp client instead of pasting to it? Something like “ claude mcp add codex-high -- codex -c model_reasoning_effort="high" -m "gpt-5-codex" mcp-s
58.
by foobar10000 1y ago
In fairness, assuming 4/8 ports in, 4 ports out, operating at some ungodly GHz using a custom GaAs or SiGe chip, and working in gearboxed scrambled space with very clever input mac prefix mapping and output scrambled/gearboxed pre
59.
by foobar10000 1y ago
Yes he very much does :)
60.
by foobar10000 1y ago
There are always DPUs like NextSilicon, NVidia's Blue*, etc - at under 100 ns of SoF to fast compute though.

More ›