Portrait of Onur Solmaz

Hey, it's Onur.

Engineer @Hugging Face 🤗, maintainer @OpenClaw 🦞.

I write about building agents, software systems, and the things I learn while shipping them.

Compact Expanded
Show all posts

Pinned

2026-07-14 Theoretical Upper Bounds for LLM Throughput
2026-07-13 Building an AI de-smeller
2026-07-30 Semantic vendoring
2026-08-27 We have AGI, but it cannot write an abstract
2026-08-25 Autoplan turns agent design into decisions
2026-08-24 Measuring model harness sensitivity
2026-08-23 Making personal training universal and free
2026-08-23 Standardized test harnesses for model benchmarks
2026-08-20 Composing agent workflows to unblock themselves
2026-08-19 Reactive agent workflows with Autoplan and Autoimplement
2026-08-10 Deterministic monitors make long-running agents reliable
2026-08-10 Local AI will favor LPDDR and MoE
2026-08-09 Theoretical limits can expose inference problems
2026-08-08 Use workflows when models cannot follow rules
2026-08-08 Be not afraid of machines, but of humans wielding the machines
2026-08-06 Semantic vendoring keeps customized tools up to date
2026-08-05 When model emotions imply model welfare
2026-08-04 Agent harnesses need session search
2026-08-04 Patterns across 720,000 agent tool calls
2026-07-31 How diffusion language models generate text
2026-07-24 Benchmarking Laguna S 2.1 on DGX Spark
2026-07-23 Open models will reward market-neutral agents
2026-07-22 Public Goods Need Room to Grow
2026-07-22 Brokerkit Makes Agent Access Explicit
2026-07-21 Report Per-Session and Aggregate Token Throughput
2026-07-20 Deterministic Agent Workflows Belong Inside the Harness
2026-07-20 When Every User Can Build Extensions
2026-07-18 AlmanBench Measures German Simplification
2026-07-18 Language Models Should Simplify English
2026-07-17 Upper Bounds for Local Model Throughput
2026-07-16 Secure Agent Access With Brokerkit
2026-07-15 Fable made Alman possible
2026-07-15 Local and open source AI has to win
2026-07-15 How to get scraped
2026-07-14 Don't give agents destructive credentials
2026-07-14 Write-up of Reiner Pope's Lecture: How GPT, Claude, and Gemini Are Actually Trained and Served
2026-07-13 Cursor has transitioned successfully to the agentic era
2026-07-05 Safe Hugging Face Access For Agents
2026-07-04 OpenAI's gamified subsidies
2026-07-02 Local AI must fit real budgets
2026-07-01 Open source AI needs cheap local speed
2026-06-29 Good READMEs say what tools are
2026-06-24 Centralized storage wins for AI artifacts
2026-06-24 Owning my X posts through my blog
2026-06-23 Local models for real-time triage
2026-06-23 Schemator reviews data models field by field
2026-06-22 Time-decayed rankings for model popularity
2026-06-22 Open model brands outlive their owners
2026-06-22 Popularity ranking cheatsheet
2026-06-20 Popularity rankings for open models
2026-06-20 Better queues for agent UIs
2026-06-17 The localening is here
2026-06-15 Gemini TTS raises the bar
2026-06-14 Making agent-written code less sloppy
2026-06-13 Specify min_iter and max_iter
2026-06-10 CLAUDE.md to AGENTS.md symlinker
2026-05-27 What clankers are NOT
2026-05-25 Keeping a workstation + terminal device is more feasible for running agents
2026-05-16 i0 to i4 interest scale
2026-05-15 Keeping an INTERESTS.md file
2026-05-15 Towards 1-click setup for local models in OpenClaw
2026-04-18 Agent blurbs for letting your agent install stuff
2026-04-15 Own your AI infrastructure
2026-04-15 People's AI
2026-04-14 Gemma 4 first impressions on OpenClaw
2026-04-10 Anthropic is optimizing for general knowledge work
2026-04-06 Introducing AI Battle
2026-04-04 A throttling protocol for model providers
2026-04-04 ACP sessions as JSONL
2026-04-03 Plainer language for better agent UX
2026-03-30 acpx v0.4 ships Agentic Workflows
2026-03-28 Token Leverage
2026-03-27 Non-dev knowledge work needs version control
2026-03-24 MCP vs CLI
2026-03-21 Agent infrastructure needs Kubernetes
2026-03-20 Merge first, review later
2026-03-20 Implementation plans keep agents on track
2026-03-15 GitHub needs extensible repo rules for agents
2026-03-11 1 to 5 agents
2026-03-08 Messaging apps can also be AI apps
2026-03-08 Using coding agents in Telegram and Discord via ACP
2026-03-08 Telegram/Discord is my IDE
2026-03-07 AI security, the lethal trifecta, and Linus's Law
2026-03-04 GitHub needs first-class support for working with agents
2026-03-03 Intelligence wants to be free
2026-03-01 Use plans to survive agent context compaction
2026-03-01 Building a static X to blog publishing flow
2026-03-01 Inference scaling can reduce coding model quality
2026-02-28 Archiving my X posts in my own blog
2026-02-28 Agentic Engineering needs rigor, not just intuition
2026-02-25 MIT licensing as the default for open source
2026-02-23 AI filmmaking quality beyond ragebait content
2026-02-23 Post-GPL philosophy for open source
2026-02-19 Family intelligence and privately owned heirloom AI
2026-02-19 AI psychosis, self-regulation, and sterile agent design
2026-02-15 We need a protocol for agent-to-app interaction
2026-02-14 SaaS must adapt to agent consumption or get replaced
2026-02-13 I built a coding agent two months before ChatGPT existed
2026-02-07 The Linux FUD playbook is repeating for AI
2026-02-06 Family AI starts with owned context and values
2026-02-03 Limits of the farming analogy for AI
2026-02-03 Local LLM demand will rise as subscriptions compound
2026-02-03 The farming analogy for AI doesn't hold up
2026-02-02 Agent etiquette is becoming an organizational necessity
2026-02-01 AI psychosis and AI hygiene
2026-01-24 Python limitations in the agent era
2026-01-22 Responsible engineering in agent workflows
2026-01-21 Agents as enforcers of engineering culture and process
2026-01-17 GitHub's trust model breaks under AI-generated code
2026-01-17 Stop defaulting to weak coding agents for serious work
2026-01-17 GitHub has to change
2026-01-12 A --skill convention for distributing agent capabilities
2026-01-11 You don't need a skill registry (for your CLI tools)
2026-01-10 Anthropic pricing discourages agent-native development
2026-01-10 Anthropic's pricing is stupid
2026-01-04 Having a "tools" repo as a developer
2026-01-03 Christmas of Agents
2025-12-31 GPT-5.2 xhigh feels like a careful systems debugger
2025-12-27 Agent progress is compounding faster than teams realize
2025-12-26 Depth on Demand
2025-12-22 How to stop AI agents from littering your codebase with Markdown files
2025-12-14 Language-agnostic interoperability layer for LLM APIs
2025-12-06 Agentic coding tools should give more control over message queueing
2025-10-13 Vibecoding this blog
2025-09-29 Google's Code Review Guidelines (GitHub Adaptation)
2025-09-08 CLAUDE.md to AGENTS.md Migration Guide
2025-08-03 Typed languages are better suited for vibecoding
2025-07-13 Workaround for Claude Code running `python` instead of `uv`
2025-07-05 Day 47 of Claude Code god mode
2025-06-04 Predictions by Anthropic Researchers
2025-05-31 SCP-3434: Istanbul Taxi Superorganism
2025-05-24 Auto-generating pull request documentation with Claude Code and GitHub Actions
2025-04-26 Working on the weekend
2025-02-26 Don't delete to fix
2025-02-21 Warmup and cooldown
2025-02-20 Satya Nadella on knowledge work
2025-01-26 Monetize AI, not the editor
2025-01-18 Calling strangers uncle and auntie
2024-12-29 AGI is what generates evolutionarily fit and novel information
2024-12-16 Our muscles will atrophy as we climb the Kardashev Scale
2024-11-02 The Kilowatt Human
2024-09-22 Python may overtake JavaScript in an AI-first workflow
2024-09-14 Open Thought > Closed Thought
2024-08-28 A failure scenario for AI infrastructure under scale
2024-08-18 Why should anyone boot *you* up?
2024-07-06 Frequencies of Definite Articles in Written vs Spoken German
2024-05-18 Stripe Subscription States
2024-05-04 Economic Burden of Language Complexity
2023-06-17 Cognitive Biases Ranked by Popularity
2023-05-11 Code-Driven Videos
2022-10-18 Microtonal Piano
2020-04-26 Take Control of Your Feeds
2019-10-21 Blockchain Fee Volatility
2019-04-20 Equilibrium in Cryptoeconomic Networks
2019-02-24 Scalable Reward Distribution with Changing Stake Sizes
2019-02-09 Bitcoin's Inflation
2018-10-18 The Anatomy of a Block Stuffing Attack
2018-08-03 Mathematics of Bonding Curves
2018-04-24 Lumped L2 Projection
2018-04-13 Disadvantages of Engineering Notation in Finite Elements
2018-04-01 Variational Formulation of Elasticity
2018-01-24 Discontinuous Divergence Theorem
2017-12-22 Nonlinear Finite Elements
2017-12-12 Coupled Finite Elements
2017-12-07 Time-Dependent Finite Elements
2017-11-21 Vectorial Finite Elements
2017-11-20 Reshaping Higher Order Linear Systems
2017-11-14 Linear Finite Elements
2017-10-12 Balance Laws
2017-09-20 Taylor and Volterra Series
2017-07-22 Isomorphisms in Linear Mappings between Vector Spaces
2017-07-13 Metrics and Natural Isomorphisms
2017-07-06 Duality of Vector Spaces
2016-04-20 Solving Constrained Linear Systems
2014-11-17 Simple recursive implementation of deCasteljau's algorithm for Bezier curves in Python