#coding-agents
48 posts
2026 - October
1 post2026 - September
10 posts- The Limits of LLMs: Good at Reproducing What Exists. Expensive for What Doesn't
- LLM Benchmark v4: Genesys PI, a New Brazilian Contender!
- LLM Benchmark v4: Opus 5.5, GPT Sol/Luna 6, Mimo 2.6, Grok 4.7
- Stop Making Excuses and Ship More! With AI, the Premise Changed. Get It.
- Talking a Bit About My AI Skills
- Why Things Like TypeSafe AI Don't Interest Me
- New LLM Benchmark v4: Retesting 39 LLMs (Part 2)
- New LLM Benchmark v4: Retesting ALL the Top LLMs (Part 1)
- AI Challenge: Converting NES ROMs to the Master System/SMS
- AI-MEMORY 2.0 - The Best Memory System for Agents and Teams
2026 - August
2 posts2026 - July
6 posts- New LLM Benchmark: I Reran Every Test!
- LLM Benchmark: Is Opus 5 Any Good?
- What's New in My AI-MEMORY: Switch AI Agents Without Losing the Session
- LLM Benchmark: Has Kimi K3 Reached Claude Opus Level?
- How Do I Protect Myself From My Agents Deleting My Stuff?
- LLM Benchmark: Sonnet 5 Fails, Gemini Flash Surprises, Sakana Fugu Almost Reaches Tier A
2026 - June
3 posts2026 - May
5 posts2026 - April
6 posts- LLM Benchmarks: Is It Worth ($$) Mixing 2 Models? (Planner + Executor)
- LLM Coding Benchmark (May 2026): DeepSeek v4, Kimi v2.6, Grok 4.3, GPT 5.5
- Clean Code for AI Agents
- LLM Benchmarks Part 2: Is It Worth Combining Multiple Models in the Same Project? Claude + GLM??
- Why LLMs Aren't Giving You the Result You Expect | Why I Prefer Claude Code Today
- VS Code Is the New Punch Card
2026 - March
2 posts2026 - February
1 post2026 - January
6 posts- AI Agents: Which One Is Best? OpenCode, Crush, Claude Code, GPT Codex, Copilot, Cursor, Windsurf, Antigravity?
- AI Agents: Is GLM 4.7 Flash really that good?
- AI Agents: Installing LSPs for Crush
- AI Agents: Comparing the Top LLMs of 2026 on the Zig Challenge
- AI Agents: Locking Down Your System
- Omarchy 3 - One of the Best Coding Agents Out There: Crush
2025 - May
2 posts2025 - April
4 posts2026 - October
1 postLLMs + Agents = competent enough
A synthesis of what I've been measuring and writing: LLMs converged. The transformer architecture got near the ceiling, the parameter war is over, and the leap everyone felt in 2025 and 2026 came from the agentic harness (tool calling, prompt caching, long sessions), not from a base intelligence orders of magnitude larger. I retell the timeline from 2019 onward with sources, explain why the agent allows a less dense model, and why that's already competent enough to automate real software engineering.
2026 - September
10 postsThe Limits of LLMs: Good at Reproducing What Exists. Expensive for What Doesn't
Every LLM was trained only on what's public, but 81.5% of GitHub activity happens in private repositories, and the source code of every proprietary software never made it into training. In my retrocomputing projects I felt firsthand where the machine stalls: with no reference and no oracle, it spins in directed brute force.
LLM Benchmark v4: Genesys PI, a New Brazilian Contender!
LUA Vision reached out on LinkedIn offering test access to Genesys PI, a Brazilian model trained from scratch. I ran both tiers, House and Enterprise, through my v4 sabotage benchmark, compared them against Opus, GPT-6, Grok and the Chinese competition, and looked at price, advantages and where it makes sense in the Brazilian market.
LLM Benchmark v4: Opus 5.5, GPT Sol/Luna 6, Mimo 2.6, Grok 4.7
Five new models entered my v4 sabotage benchmark: Opus 5.5, GPT 6 sol and luna, MiMo V2.6 Pro, and Grok 4.7. I compare against external benchmarks, answer whether upgrading is worth it, and Grok 4.7 regresses badly against its own predecessor.
Stop Making Excuses and Ship More! With AI, the Premise Changed. Get It.
With frontier LLMs making fewer mistakes than most human devs, the only real excuse left for holding back a deploy is your own review process's bureaucracy. I use my open source projects, cross-platform CI/CD, and real contribution numbers to show how I've been doing this since February.
Talking a Bit About My AI Skills
Agent skills are just prompts in text files. I keep 25 in my my-skills repo, from pr-audit to fact-check, and they have already closed over a thousand PRs and issues across my projects. The recommendation: don't copy somebody else's skill, build your own.
Why Things Like TypeSafe AI Don't Interest Me
A handful of people showed up in the comments asking 'have you tried TypeSafe?'. I use that as a hook to lay out my rule about new AI tools, actually investigate the case, and explain why I don't waste time optimizing token spend.
New LLM Benchmark v4: Retesting 39 LLMs (Part 2)
The full v4 table, nearly forty models ranked by vigilance, not completeness: six ties at the top, a free model beating half of Claude, GLM winning on a flat-rate plan, and the test I ran to find out whether Kimi or DeepSeek are secretly hiding a connection to Claude under the hood.
New LLM Benchmark v4: Retesting ALL the Top LLMs (Part 1)
I rebuilt my benchmark from scratch for the third time. I threw out an entire version, built a Rails app sabotaged with real CVEs, and spent over $4,000 in 9 days. In this Part 1: the process, the cost, and the why.
AI Challenge: Converting NES ROMs to the Master System/SMS
I tried using frontier LLMs to convert NES ROMs into Master System games. I explain the architecture, why translating 6502 to Z80 runs slow, the emulator as a feedback oracle, and where the mappers stall everything.
AI-MEMORY 2.0 - The Best Memory System for Agents and Teams
ai-memory hit version 2.0 with the open OKF format, local embeddings on by default, and support for teams working in parallel. I compare it with the competition and show what only ai-memory delivers.
2026 - August
2 postsLLM Benchmarks: The latest Deepseek v4, stop asking
I ran the new Deepseek v4 snapshots through the benchmark: Flash jumped from 80 to 90 and Pro from 82 to 91, both Tier A, the cheapest in the pack. I still prefer Kimi K3 and GLM 5.3, and I explain why a close score does not mean an equivalent model.
LLM Benchmarks: Qwen 3.8, GLM 5.3, Gemini 3.7, Grok 4.6
GLM 5.3 scored 94 and closed in on the leading trio. Qwen 3.8 Max jumped from 51 to 92 after it stopped hallucinating the RubyLLM API. Gemini 3.7 Flash scored 93 — after its first run was voided for cheating. And the local 27B Qwen showed that context is now the bottleneck.
2026 - July
6 postsNew LLM Benchmark: I Reran Every Test!
I reran the LLM Coding Benchmark with a harder test, three phases, native harnesses, and new tiers. Fable led, Terra tied Sol as the best GPT, and 24 models proved suitable for programming.
LLM Benchmark: Is Opus 5 Any Good?
Opus 5 scored 95/100 on the Rails benchmark, tying Opus 4.8 and edging Fable 5 by one point. Its API rate is half Fable’s, but this narrow test does not define the best LLM.
What's New in My AI-MEMORY: Switch AI Agents Without Losing the Session
ai-memory run keeps the same programming session while switching among Claude Code, Codex, and other harnesses, with searchable workstreams and integration with ai-jail and ai-usagebar.
LLM Benchmark: Has Kimi K3 Reached Claude Opus Level?
I had Kimi K3 build a chat in Rails 8 on its own: it scored 89/A, beat Opus 4.6, and was a cheaper alternative to Opus 4.8. It still lagged behind in architecture, hardening, and tests.
How Do I Protect Myself From My Agents Deleting My Stuff?
LLMs deleting files from famous people made headlines this week. In five months of heavy use, in YOLO mode, it never happened to me. But I don't trust them either: BTRFS snapshots, restic backups, sandboxing and discipline.
LLM Benchmark: Sonnet 5 Fails, Gemini Flash Surprises, Sakana Fugu Almost Reaches Tier A
In the coding benchmark, Sonnet 5 scored 58/100 after calling a nonexistent API, while Gemini 3.5 Flash scored 93/100. Sakana Fugu Ultra landed at 79/100, close to Tier A.
2026 - June
3 postsai-memory: long-term memory (Karpathy Wiki) and self-improvement (Hermes) for your projects
ai-memory turns agent sessions into a searchable Markdown wiki with hooks, MCP, and handoffs between tools. A Hermes-inspired loop promotes memories with validation, evidence, and optional review.
AI Controversy in Open Source Project Contributions - My Take
AI in open source contributions is here to stay, even with AI slop, regressions, and false bugs. The practical answer is to automate triage and auditing without taking the final decision away from humans.
LLM Benchmarks - An Update on Grok 4.3, MiniMax v3, and Opus 4.8
In the Rails 8 benchmark, Opus 4.8 keeps the lead at 95/100, while Grok 4.3 and MiniMax M3 finally become usable but remain in Tier B, behind GPT and Opus.
2026 - May
5 postsFirst Impressions Using Oh-My-Pi and OpenCode
Oh-My-Pi is more flexible for projects with varied files and artifacts, while OpenCode is more polished for ordinary code. The test showed that no harness audits everything, and Codex remains behind.
I Built a Memory System for Coding Agents: ai-memory
After finding reindexing, data loss, and broken hooks in agentmemory, Akita built ai-memory in Rust with Markdown, SQLite/FTS5, automatic capture, and agent handoff. It’s still beta.
AI Agent Memory: Karpathy LLM Wiki and agentmemory in Practice
Akita compares compaction in Claude Code, Codex, and opencode, explains Karpathy’s LLM Wiki, then tests agentmemory before dropping it over structural bugs and pointing to ai-memory.
Wrapping Up My AI Marathon: Success or Failure?
After more than 500 hours across 24 repositories, Akita ranks his experiments and concludes Codex and Claude Code support daily coding, but without review and XP, productivity turns into slop.
LLM Benchmarks: DeepSeek Unlocked! Use DeepClaude
I swapped opencode’s harness for Claude Code’s loop through DeepClaude, and DeepSeek V4 Pro rose from 69/B to 89/A in 18 minutes and $3.14. The controller’s multi-turn bug remains.
2026 - April
6 postsLLM Benchmarks: Is It Worth ($$) Mixing 2 Models? (Planner + Executor)
Three rounds show multi-agent doesn’t beat solo Opus 4.7 in opencode, which scored 97/100 in 18 minutes for about $4. GPT 5.4 xHigh with a medium executor saves money but loses 3 points.
LLM Coding Benchmark (May 2026): DeepSeek v4, Kimi v2.6, Grok 4.3, GPT 5.5
I reaudited 24 LLMs on the same Rails app with RubyLLM: Opus 4.7 and GPT 5.4 tie at 97, GPT 5.5 reaches 96, Kimi and Gemini become Tier A, and DeepSeek V4 Pro reaches Tier A only through DeepClaude.
Clean Code for AI Agents
Clean Code carries different weight when an agent is the primary reader: small code, greppable names, provenance context, headless tests, and explicit rules reduce navigation, cost, and errors.
LLM Benchmarks Part 2: Is It Worth Combining Multiple Models in the Same Project? Claude + GLM??
I tested seven combinations in Claude Code, opencode, and Codex, but no sub-agent was called. Variance across runs and harnesses keeps solo Opus as the default for greenfield projects.
Why LLMs Aren't Giving You the Result You Expect | Why I Prefer Claude Code Today
Every week somebody tells me 'I canceled my Claude plan, it just doesn't perform as well as GPT for me.' I've got 500+ hours in Claude Code and Codex, 400k lines generated, and both deliver. The difference isn't the model. It's how people are talking to it.
VS Code Is the New Punch Card
In the era of coding agents, typing everything by hand in VS Code is becoming the modern equivalent of a punch card. Foundations did not become legacy.
2026 - March
2 postsClaude Code's Source Code Leaked. Here's What We Found Inside.
I analyzed Claude Code’s leaked source map and found hidden features, layered memory, multi-agents, and DRM based on xxHash64. The code also exposed a product that’s difficult to maintain.
ai-jail: Sandbox for AI Agents — From Shell Script to Real Tool
ai-jail is a Rust tool that wraps bubblewrap (Linux) and sandbox-exec (macOS) to safely run AI coding agents like Claude Code, Codex, OpenCode and Crush in a sandbox.
2026 - February
1 postAI Agents: What Would Be the Best Programming Language for LLMs?
A thought experiment on what a programming language designed for LLM coding agents, rather than human programmers, would look like.
2026 - January
6 postsAI Agents: Which One Is Best? OpenCode, Crush, Claude Code, GPT Codex, Copilot, Cursor, Windsurf, Antigravity?
A practical take on the current state of AI coding agents, why proprietary harnesses matter, and why I am sticking with Crush.
AI Agents: Is GLM 4.7 Flash really that good?
Running GLM 4.7 Flash locally on an RTX 5090 via Ollama to see if it can finally match commercial LLMs on a hard coding challenge.
AI Agents: Installing LSPs for Crush
My Crush config with LSPs wired in, plus an Arch Linux script that auto-detects installed languages and installs the matching language servers.
AI Agents: Comparing the Top LLMs of 2026 on the Zig Challenge
I put the top commercial and open source LLMs of early 2026 through a nasty Zig + llama.cpp migration challenge to see which ones actually deliver.
AI Agents: Locking Down Your System
I show how to isolate coding agents with Bubblewrap: the jail leaves only the project writable, hides the rest of the system, and mounts only what is needed, reducing the risk of destructive commands.
Omarchy 3 - One of the Best Coding Agents Out There: Crush
I used Crush with Claude Opus 4.5 on a Hugo blog and a Zig chat using llama.cpp, CUDA, and Qwen3. It refactored scripts, updated APIs, fixed a segmentation fault, and delivered a working binary for nearly $8.
2025 - May
2 postsRANT - LLMs are LOOT BOXES!
Why LLMs for real coding behave like gacha loot boxes, and why every incentive in the industry pushes you to burn more tokens.
When Do LLMs Fail at Programming? A More Realistic Use Case.
A real-world attempt to build a Zig program using llama.cpp with Gemini 2.5 Pro and Aider, showing where LLMs crumble on new languages and libraries.
2025 - April
4 postsTesting the Newly Released Open Source LLM - Qwen3 (with Aider and Ollama)
First hands-on impressions of Qwen3 running locally via Ollama and remotely via OpenRouter, tested with Aider on real refactoring tasks.
Testing LLMs with Aider on RunPod - which one to use for code?
Hands-on testing of Qwen2.5 Coder, DeepSeek, Codestral and others with Aider, locally on an RTX 4090 and on RunPod, comparing them against Claude and Gemini.
Your Own Free Universal Co-Pilot Running Local: AIDER-OLLAMA-QWEN
How to run Aider with Ollama and Qwen 2.5 Coder locally to build a free, universal Co-Pilot alternative that works with any editor.
LLM Hello World: Building Your Own Local AI Chat
A hands-on Hello World experiment explaining how LLMs actually work under the hood, building a tiny local chat CLI with Qwen 2.5 Coder running on a single GPU.