Hey hey, Useful AI coding links to read/watch for this week of March 18, 2026.
If you want to get such list every Wednesday, subscribe via the form at the bottom.
For Premium Members
I Tried 3 LLMs for Visual Design: Is Gemini Still The Best?
A 12-minute video for Paid subscribers.
From My YouTube channel
I Tried to Use the NEW 1M Context Window in Opus 4.6
Anthropic announced the general availability of 1M Context for Max plans, so I tried it out.
Stop Trusting AI-Generated Tests Blindly: My Examples
I noticed that AI agents can generate automated tests, but the problem is that we can't fully trust them.
I Tried NEW /voice in Claude Code vs Wispr Flow: Which is More Accurate?
There's a new /voice command for dictation in Claude Code. I tried it out against Wispr Flow that I had been using for months.
Fixing Bugs with AI Agents: "The Right Way"
I saw this tweet about fundamental first principles of bug fixing, and let me show it in action.
Codex GPT-5.4 Fixed A Bug That Opus 4.6 Missed
Quick 2-minute video
News About Codex
Subagents are now available in Codex.
x.com
You can accelerate your workflow by spinning up specialized agents to: • Keep your main context window clean • Tackle different parts of a task in parallel • Steer individual agents as work unfolds
How to run subagents in Codex
x.com
Long article on X by Emanuele Di Pietro.
Library of 425+ custom agent roles that you can drop directly into Codex.
x.com
An enormous library of 425+ custom agent roles that you can drop directly into Codex. Subject matter experts of all kinds, curated from across Github.
Codex just shipped updated 0.114.0 with one of its most requested features ever: hooks.
x.com
But..there's a catch. (high level dad joke right there) There are only two hooks right now: SessionStart and Stop.
Sam Altman: Usage of Codex is growing very fast
x.com
The Codex team are hardcore builders and it really comes through in what they create. No surprise all the hardcore builders I know have switched to Codex.
News About Claude Code
1 million context window: Now generally available for Claude Opus 4.6 and Claude Sonnet 4.6.
x.com
Opus 4.6 scores 78.3% on MRCR v2 at 1 million tokens, highest among frontier models. Load entire codebases, large document sets, and long-running agents. Media limits expand to 600 images or PDF pages per request.
We doubled Claude usage on weekends, and outside 5–11am PT on weekdays for the next 2 weeks.
x.com
Lessons from Building Claude Code: How We Use Skills
x.com
Big article on X by Thariq from Claude Code.
You can now add a setup script in Claude Code on the web and desktop
x.com
Use these to automate setup before Claude Code launches on a cloud environment. It’s particularly useful for installing dependencies, settings, configs, etc.
You can use /model opusplan
x.com
Then Claude Code uses Opus for planning and Sonnet for execution. This is effiencymaxxing.
VoltAgent/awesome-agent-skills
github.com
Claude Code Skills and 500+ agent skills from official dev teams and the community, compatible with Codex, Antigravity, Gemini CLI, Cursor and others.
That's it for this week, see you in the next newsletter issue!
Again, if you wanna get this to your inbox every Wednesday, subscribe via the form below.
