Dory Zidon: AI Engineer & Researcher

AI in engineering

Deep dives into AI tools, workflows,
and best practices for engineers

26/10/26

Anthropic's AI-Native SDLC Guide, Explained Stage by Stage

18/10/26

Kanban, Stories, Features, Sprint: Are These Planning Methods Even Relevant With AI?

9/5/26

Karpathy vs 8 CLAUDE.MDs: head-to-head

24/4/26

Workflows That Ship: I Tested 3 AI Orchestration Patterns Head-to-Head

25/8/26

Become a 10X AI Engineer

4/8/26

Interviewing AI Engineers: Practical Questions And Exercises

6/7/26

Reviewing AI PRs at Scale: 5 Levers to Stop Drowning

28/5/26

Sandboxes: when CEOs Ship code

26/3/26

Self-Healing AI Dev Workflow

19/2/26

AGENTS.MD / CLAUDE.MD Layering

11/3/26

OpenSource AI Code Reviewer

26/2/26

AGENTS.MD / CLAUDE.MD That Actually Work

8/1/26

Casino Code Part 2

8/1/26

Casino Code Part 1

18/12/25

AI Will Replace You, Let's build what's next.

Build in public

Building, learning, and sharing
experiments in how I orchestrate.

October 26, 2026

Stop Approving Plans You Didn't Read

October 26, 2026

Protect Your Context. Kill MCP. Use Agent Tools.

February 14, 2026

AI Orchestration - Prove - What the AI team Delivered

February 14, 2026

AI Orchestration - Deploy: Agents in Action

February 14, 2026

AI Orchestration - Build: Team of AI Agents from Scratch

Experiments

Research & Experiments Results
& Interactive Conclusions

Karpathy vs 8 CLAUDE.MDs

9 variants · 126 builds · 378 judgements · 27 smoke tests

Karpathy's CLAUDE.md vs the internet's most-liked CLAUDE.md / AGENTS.md files, plus a few of my own. The results aren't what you'd expect.

Understanding Is All You Need

14 research passes · ~200 agents · 24 numbers · every figure sourced and dated

AI does blast out code, and that part is real. Twenty four numbers on what it costs: 17x the code for 1.3x the software, review time up 441 percent, and zero studies measuring what it does to the people writing it.

Kanban, Stories, Features, Sprint? Which Planning Method Works Best With AI?

Kanban, stories, features, sprint · 2 agents · 50 builds

Are old software planning methods still relevant with AI? One app idea, planned five ways from a bare goal to a full sprint, built 50 times by Claude Code and Codex. No method won outright, and the sprint plan made Claude build less.

Blog

October 26, 2026

Anthropic's AI-Native SDLC Guide, Explained Stage by Stage

Anthropic's AI-native SDLC playbook, walked through stage by stage: what changes, the steps, a real example file, and where to start.

Read

October 16, 2026

Kanban, Stories, Features, Sprint? Which Planning Method Works Best With AI?

I had Claude Code and Codex each turn one app idea into five plans, from a bare goal to a full sprint, and build every plan five times. Fifty builds later, no method won, and the lessons were about alignment and consistency.

Read

See all posts

Get the latest updates

Thank you! You have subscribed!

Oops! Please wait a few minutes and submit again.

D0ry Z1don

[AI Engineer. Researcher. Builder. Orchestrator.]

[I've spent 20+ years building software, and the last few years researching how AI is fundamentally changing how we do it.
From LLM infrastructure at Martian to building AI agent orchestration systems, I work at the intersection of AI and engineering every day. O-1 visa holder recognized
for extraordinary ability in AI engineering.
I publish what I learn openly.]