I now start every new piece of work with the same steps in Codex:
1. Pop open voice mode
2. Brain dump what I’m trying to accomplish
3. Ask it to pull in relevant context from Slack, Fathom, Gmail, and Notion
4. Ask it to propose a build order for me to review and ask any clarifying questions it needs
This has dramatically reduced the “activation energy” required for me to tackle anything I need to work on.
And as a result, I procrastinate less and stay in flow for much longer periods of time.
I walk through the full process here:
https://t.co/8ufe5wbC5N
I spent over 1,000 hours learning to write effectively— because college completely failed to teach me.
Here's everything I learned distilled into 11 tweets (that you can start applying today):
First — why the "pick one AI" era is already over, and why most people haven't realized it.
In 2024, the question was: "Which AI should I use — ChatGPT or Claude?"
In 2026, that question is obsolete — because no single AI is best at everything, and the gap between their strengths is growing wider, not narrower.
GPT-5 leads at: code generation, image creation and editing, vision tasks, and multimodal reasoning. Claude leads at: long-form writing, nuanced analysis, careful reasoning, document comprehension, and following complex instructions. Gemini leads at: web-grounded research, processing massive documents (1M+ token context), and integration with Google's ecosystem. Llama 3.3 leads at: being free, running locally, and delivering 80-90% of paid model quality at $0. Mistral leads at: multilingual tasks and efficient performance at lower cost. DeepSeek leads at: math, science, and coding at a fraction of GPT's price.
Using one AI for everything is like using one tool for every job. ChatGPT is a hammer — great for nails, terrible for screws. Claude is a screwdriver — great for precision, less useful for brute force. Gemini is a power drill — great for heavy lifting, overkill for hanging a picture.
Most people chose a hammer in 2023 and have been using it for every task since — paying $20/month for a single tool while better tools for specific tasks sit unused on other platforms.
"The question isn't 'which AI is best.' The question is 'which AI is best for this specific task.' That question always has an answer — and it's rarely the same model twice. OpenRouter lets you ask the right question."
Apple just mass-killed 9 paid apps with one free software update.
iOS 27 shipped September 22. Most people saw a new look and moved on.
A developer who tracks App Store economics looked at the feature list and said: "Apple just made $372/year in paid apps completely unnecessary. They rebuilt the functionality of 9 subscription apps and buried it inside a free update. Most people will keep paying because they'll never find the settings."
He showed me every feature and the paid app each one kills.
Here's the full list 🧵
Google just dropped a 1-hour course on agentic engineering from scratch:
00:00 – How to build your first AI agent
08:24 – Build agent memory (short, persistent, long)
28:34 – Agentic loops, long-running AI agents
40:04 – How to build MCP (MCP vs API)
1:00:22 – Multi-agentic systems
This 1-hour watch will replace 10 paid agentic courses on the internet.
Bookmark this. Watch this weekend.
🚨BREAKING: Claude has a mode called infinite personal tutor.
It can teach you any skill from zero in 4 hours.
Here are the 09 prompts to learn anything:
(Pick a skill. Save the prompts. Start with #1.) 📌