Posts tagged "ai"
-
AI Problems Are Just Human Problems Amplified
AI fatigue, perfectionism vs non-determinism, and garbage in/garbage out. These are not new problems. AI just makes them obvious faster.
-
llm-mux: Why I Rebuilt Lok
Lok got 300+ installs, then I rewrote it. The abstractions were wrong. llm-mux has roles, teams, and proper apply/verify. Here's why.
-
finna: Multi-Model Debate, Spec, and Implement
A standalone tool that takes an idea, debates it across Claude, Codex, and Gemini, creates a roadmap, writes specs, and implements. Planning and execution in one pipeline.
-
Building a JSON Parser with Multi-LLM Orchestration (Part 1)
Using lok to orchestrate four LLMs debating design decisions, then synthesizing specs for a Rust JSON parser. The debate phase surfaced edge cases no single model would have caught.
-
Lok Part 5: Multi-Agent Planning with lok spec
Lok gains a spec command that turns task descriptions into structured implementation plans. Multiple LLMs propose, debate, and converge on a roadmap before any code gets written.
-
ARF: Structured Reasoning for AI Agents
Moving beyond chat prompts to structured agent communication. Why unstructured data lets LLMs run wild, and how ARF enforces what/why/how before acting.
-
I Built a Robot to Help Me Understand People
I'm good with machines but bad with people. So I built a tool that reads what my coworkers write and helps me understand them better.
-
Lok Part 4: The Self-Healing Loop
Lok gains agentic workflows, fixes its own bugs, and finds a real bug in Discourse that I just pushed upstream.
-
Lok Part 3: Dogfooding and Code Review
PR review, codebase explanation, and lok opening 25 GitHub issues on itself. Plus parallel workflows and context detection.
-
Lok Part 2: Workflows and Local LLMs
Declarative multi-step pipelines and Ollama integration. The workflow system that's also a plugin system.
-
I Don't Type Every Word You Read
Most modern work is already mediated by machines. Writing is just late to admit it. On delegating friction, not thinking.
-
Introducing Lok: A Local Multi-LLM Orchestration Control Plane
When one AI isn't enough. Lok is a CLI tool that coordinates multiple LLM backends, routing tasks to the right model and letting them debate each other.
-
The Unholy Trinity: Nix Shells, SSH Config, and Claude Code
How proper environment setup turns AI coding assistants from novelty into genuine infrastructure. Nix makes everything reproducible. SSH makes everything reachable. Claude ties it together.
-
Letting AI Pick the Project
An experiment in creative delegation: I asked Claude to build whatever it wanted. It chose to build a tool it wished existed.
-
Building LLM-TUI: Never Lose Context Again
Built a terminal AI chat that remembers everything. File context persists across sessions, tools execute with confirmation, and you never lose your place.
-
Like a Pig in Shit: Why I Love AI "Slop"
I'm having the time of my life building projects with AI assistance. Call it slop if you want - I'm too busy shipping to care.
-
How I'm Using Claude Code for Daily Development Work
Real-world AI pair programming: what works, what doesn't, and how to avoid the traps