Onur Solmaz blog

I was not expecting this to receive so much attention!

sol really loves to use words like envelope, capsule, receipt

I have performed a distillation attack on my PERSONAL TRAINER Since more than 2.5 years, I have recorded over 300 of our sessions, each up to 90 minutes long Why would I do that?

We need to normalize measuring and judging models against a standardized test harness "Oh but model X performs best in their own proprietary harness" I could not care less.

This actually makes the most sense to me from what I've read till now.

my @pidotdev extensions give me so much joy. so much tedium that I automated away, that I know are guaranteed to be performed the way I want, unlike with pure skill files

Who is actually training a model to come up with better headlines/thumbnails?

What a time to be alive @codex_reset

This is your daily reminder that you can just vibe your dream terminal dashboards with @ratatui_rs and it just works amazingly

Once nice thing with workflow graphs is to be able to compose them I stitched `autoplan` + `autoimplement` workflows into `monitor` in an "issue detected" path: monitor -> wait 30min -> issue detected -> autoplan -> autoimplement -> monitor Previously, mo…

We live in a simulation

God save my soul I might have built quite a bit of a rube goldberg machine here in pi My "autoimplement" skill of 6 months is now an actual graph based workflow I even have "autoplan".

Before someone says it, I know https://t.co/MYmBWy3vOo exists.

A pi-based alternative to codex review, for your auto-review loops It is like codex review, but it works with all the models that pi supports, including gpt 5.6 https://t.co/vDszMNytBK

It seems my tweet ragebaited some unintentionally (best kind of ragebait, with plausible deniability) My point here was that, you would be sending a compressed version of human knowledge back in time Having compressed such high intelligence per parameter,…

POV: You've automated monitoring your agents I did a lot of quality-of-life improvements to osolmaz/pi-workflows.

before anyone thinks anything, I am using codex's remote compaction in pi btw

Qwen 3.8 27b weights are probably the most civilization-changing 20gb of data published, until now Imagine sending those 20gbs back in time, to 2006

The only reason I'm using Sol at this point is that it's subsidized and fast tbh

How insane is tailscale/wireguard Discord is blocked in turkey, so I just set my spark as an exit node.

People are going nuts over qwen 3.8 27b and it's justified I'm personally waiting for qwen3.8-35b-a3b.

.

Qwen 3.8 27b just scored 4% higher than GLM 5.2 in a non-trivial private benchmark I am working on Not sure if there is an issue with the benchmark or the new 27b is that good

Me: I want token subsidies Mom: We have token subsidies at home

Lisan Al Gaib

Sam Altman repeated many times that the ideal model for him is a model that does not have a lot of world knowledge, is small by comparison and very cheap to run, and can reconstruct knowledge via reasoning as need arises A paradox with that statement is t…

HOW IS THIS MODEL BETTER THAN GPT 5.6 SOL AT WEB DESIGN

Lol now think how 27b would compare to the quantized opus 4.6 anthropic served us, instead of the full precision one reported in the benchmark Too bad they can't quantize our 27b

Get your GGUFs and NVFP4s here beloved Spark users

me: I want Opus 4.6 mom: we have Opus 4.6 at home Opus 4.6 at home:

Spatiotemporal composability???

https://t.co/lGFGoCNnWn fixes this

😎 https://t.co/xxfnJQ9MN3

A similar thing can be done with codex session compaction summaries as well btw I had verified the other day that a compaction summary created by one account can be used by another account if it has the encrypted blob ie. they were not keyed/guarded with…

This type of estimations are useful not only for local but for all inference providers Get a ballpark of max possible throughput for a model, directly calculate your revenue

mfw trying not to think about the 100 models that dropped this week I’ll be offline 1 week to touch grass and reset I’ve literally not given any break since claude code came out lol let’s see if I can resist the urge to check twitter

I've felt the lack of a certain feature of codex desktop app since I went back to the CLI: scheduled tasks Codex desktop app can keep track of a task until it is properly finished.

Apparently this went to HN front page briefly last night, so sharing it here as well YOLO safely with your agents 🤖 Give your GitHub/Hugging Face accounts + ability to run sudo safely to your agent.

This.

urgh typo, I meant 35-40 tok/s

Apparently Alibaba did all this work, published a paper, but did not create a public backup of repos used in the review tasks https://t.co/aoZoYD79Vh Then keycloak and nodejs repos got force pushed, so the commits for 6 of the tasks got lost :( I recovered

I have made an update to my theoretical upper bound calculation to also predict prefill speed Prefill relaxes the assumption we make for decode, that it is only be memory bottlenecked.

You can use the calculator in https://t.co/OapFVFtVB6 now to predict theoretical upper bounds from https://t.co/tNttYLI8Ih 🙌 A little note, pay attention to the speculative decoding coefficient rho.

Rules that current models cannot follow with skill, i.e. a single prompt: Chasing P2 and P3 errors: I have a rule in my autoimplement skill to stop reviewing once the last round of review only generates P2 errors or less.

I'm running some private benchmarks on @liquidai's LFM 2.5 2.6B, and if my results are correct, we might have a new champion for <10b category Scores significantly higher than gemma 4 e4b, which is 3-4x its size

Anyone else use Alibaba’s AACR bench for code review evaluation?

mainstream got extremely scared when moltbook went viral. but it was just a worthless marketing stunt why?

Obligatory repost of “Clippy” by Gwern https://t.co/UfQwH8NfbV

Super curious what second hand price gb10's will converge on

Super excited for https://t.co/VYuTSdkbVP!