Pinned Tweet
notes from this week...
- Interviewed 5 founders/business owners. Three were struggling with AI transformation. The major areas = shared context and workflows. Nothing was sticking, which sounds like a skills and adoption problem, not a tooling one.
- The other two were trying to scale GTM: one with zero paid customers, the other with early pilots but no PMF or clear ICP. Growth dollars exist after understanding customers, positioning, and dialing in product.
- Four of the five felt behind on AI adoption. They want to go deeper but have no idea how to budget for it. “What token formula per employee should we use?”
- Friends who code for a living aren’t writing code by hand anymore. They review, and change what needs changing.
- All my builder friends are diverging on the word “focus.” Some are doubling down on the AI spray and pray game. Going wide, building literally anything and everything. Others are killing every sidequest to pour everything into one product / business... to be a missionary !!
unlearning 20 years of building product to become a “native ai builder”
things i’m trying and really enjoying
- building a cloud based agentic workforce.. goal is to reach or exhaust 1b high quality tokens per day
- design.. entering peak play era 😍
- building context graphs
- wide deep dive research
- talking to new non-technical builders who are making magic for the first time
- revisiting old projects, products, and problem statements
things i’m failing at
- context debt (project-based)
- unblocking runners and keeping agents in flight (peak 😤 when I see runners in idle for 12 hours)
- saying no to new features or products (esp during design explorations)
- keeping token spend balanced throughout the day and ramping output during off-peak hours
human craft ftw 👌
"RENDER ASUNDER"
11 x 17 in.
A limited number of riso (and non-riso) prints available here: nabihaider.com
Kenny Grant retweeted
came back to Claude after an 8h57m task and it was stuck waiting for permission to access Google Drive
ai kills the idea maze and the stress tests that happen in the 0 to 1 process
the idea maze gets built and refined subconsciously when you write the prd, the deck, the 1 page north star doc, the wireframe, the db schema etc
when you skip those, you sometimes you get lucky.. a lot of the time ai takes you down the wrong maze
Kenny Grant retweeted
We're going to look back at how much time people spent learning to use software and wonder why we thought that was normal.


















