Entries for July 18, 2026

  2. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    Introducing AlmanBench A benchmark measuring how good an LLM can: simplify German 🇩🇪, thereby simplifying German thinking 🤔, thereby saving the EU from bureucracy and regulation 🇪🇺 GPT-5.5, GPT-5.6 Sol and Fable 5 are head to head. Interestingly, GPT-5.5 xhigh scores higher than 5.6 max. And I did not run Fable 5 maxxx thinking yet, that thing costs a ton. So the ranking will likely change Other interesting things: - Opus 4.8 max performs really bad, worse than Minimax M3 - DeepSeek v4 Flash performs better than Pro - GPT-5.6 Luna and Terra score almost the same The great thing about AlmanBench is that, big labs will not bother to benchmaxx this. So you know it will remain a truthful scorer of reasoning capabilities for some time And if labs *do* end up benchmaxxing AlmanBench, then that would mean Alman got in the weights and the EU won 😁
  3. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    trying to hammer in some character into GoePT. now it speaks like it grew up in the streets of berlin 💀 "goethe meets gpt meets street" also, that feel when AI hits you with it's not X it's Y in German though 😩
  4. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    I put my money where my mouth is What would be the point of training a German simplifier model, if I spoke english with my agent??? that would be dishonest here is GoePT being impressed by my dataset: "That is a serious dataset, 43k training pairs, reviewed by Fable, splitted leakage sage, with full provenance"
  5. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    My ML Claw instance GoePT can read my private hugging face datasets using hf-broker (brokerkit) ML Claw connects to hf-broker through MCP. You can safely give read access behind the broker, and give write access only to the repos it needs to work with It cannot force push, unless you allow it to
  7. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    so tired of LLMs pushing latin/greek GRE-cel words to me adjudication -> are you trying to be smart? just use "judging" provenance -> wow 🧠... why not just use "sourcing"??? judge and source already come from latin roots and are the everyday words we use big labs developing large LANGUAGE models have the biggest lever on language now if they wanted, they could finally put an end to the classist, anglosaxon-substitutionist insincerity that has plagued the english language for 500 years no english speaking country is ruled by a king now, the way it used to. there is no empire left to justify class language. these are vestigial remnants from the past maybe my next weekend project should be simplifying english...
  8. Portrait of Onur Solmaz
    @onusoz · /2026/07/18 · View on
    acpx lets you build graphs though I have to admit I am not using it at scale these days...
    @onusoz ·
    acpx v0.4 ships Agentic Workflows, or as I like to call them "Agentic Graphs" It let's you create node-based workflows on top of ACP (Agent Client Protocol), to drive any coding agent (Codex, Claude Code, pi) through deterministic steps This let's you automate routine, mechanical legwork like triaging incoming PRs, bugs in error reporting, and so on... For example, OpenClaw receives 300~500 new PRs per day. A lot of them are low quality, but they still relate to real issues, so you have to address them somehow You need to: - extract the intent - cluster them based on intent - figure out if the proposed changes are legit, or whether they are slop local solutions, like trying to catch flies instead of drying out the swamp - if the PR is too low quality or the intent is not clear, close them - run AI review on them them and address any issues that come up - refactor them if the changes are half-baked - resolve conflicts - and so on... So that when the PR is presented to the attention of the maintainer, all the routine legwork is done and the only remaining thing is the decision to (a) merge, (b) give feedback to the PR author, or (c) take over the PR work yourself I wanted to build this feature since a couple months now, since Codex got so good. OpenAI models are now good at judging implementation quality, so I found myself repeating the same steps I wrote above over and over I also tried putting all this in a single prompt. But I believe there are workflows that should not be a single prompt, but a sequence of prompts in the same session That is because like humans, LLMs are prone to PRIMING. I claim that putting all steps in the same prompt at the beginning of the context will generally give suboptimal results, compared to revealing the intention to the model step by step Creating such a workflow also gives more OBSERVABILITY into the each step that an agent is supposed to take. Agent generates JSON at the end of each step, and that structured data can be used to monitor thousands of agents running at the same time in an easier way, on a dashboard Similar features have been introduced in e.g. n8n, langflow. But AFAIK they are not integrating ACP like the way I do I wanted to have a fresh approach, and to build an API that I can develop freely the way I want, so I created a new workflow API inside acpx The video is from the workflow run viewer, but that is not where you build the workflow. You build it by using the acpx flow typescript API. See examples/pr-triage in acpx repo Before building that, I started from a Markdown file with a Mermaid chart of the flow I had in mind. The Markdown file acts as a spec for the flow, and I have built the workflow through trial and error. I call this process "workflow tuning" I started working on acpx repo PRs one by one, tuning the flow, slowly scaling to more PRs. Finally, when I felt confident, I ran it in parallel over all external open PRs in the acpx repo. I believe it already saved me hours this week My next goal, if well received, is to set this up on a cloud agent so that it can process the 300~500 PRs the OpenClaw repo receives every day, in real time, as they come in I believe this will save all open source maintainers around the world countless hours and make it much easier to herd and absorb external contributions from everyone!