
Would love to have a quick chat with anyone building with AI about what analytics/product monitoring tools they're using and upsides/downsides (will publish results in a blog post) Shoot me a DM if you're up for a quick chat!

Product & growth. 6 yrs at Unito: pricing, activation, experiments (ASP & LTV ~5x). Building seotakeoff.com + getsensei.io. Open to AI product roles 🇨🇦

Would love to have a quick chat with anyone building with AI about what analytics/product monitoring tools they're using and upsides/downsides (will publish results in a blog post) Shoot me a DM if you're up for a quick chat!

I had been on @OpenAI almost completely for the last several months, particularly after the whole third-party harness thing, but also because of the models Have to say, I'm back with @AnthropicAI now, their latest models are exceptional, and seem much smarter than OpenAI's latest

The real world building project also continues to progress, we’ve got trusses going up, concrete poured for deck, entry and storage areas, all before roofing gets installed Aug 10

Been split between real building and software building, but shipping lots of improvements and new features for @yourseotakeoff lately Including new messaging/home page, support widget for chat, some real-world examples of growth, more product screenshots

Awesome story, and would love to see more work like this across the Maritimes!
This tweet is unavailable

Update on the real life building project: office is nearly done for major work, next steps will be finishing the interior and getting it all set up Have to backfill/do some exterior cleanup too

Just published a new update on the cottage builds for those curious: diyecohomes.substack.com/p/b…

Okay so update on how I'm using the latest vibe coding/agent tools: • Been using @NousResearch Hermes instead of @openclaw for a couple weeks • Now using @discord over @telegram whenever possible: threading makes a huge organization difference • @OpenAI is my primary driver now, design was it's weakness compared to @claudeai but I've gradually been teaching it (much better code in my experience) • Have been using desktop Codex the most lately, occasionally @cursor_ai @claudeai @conductor_build • Pushing to remove myself as the bottleneck is still the biggest battle • Still get GPT-5.5 and Opus 4.7 to go back-and-forth on specs and implementation plans until they agree, then can get whatever to implement it

Great piece, the opening line says it all
Start projects, not companies (at first)


This week's version of my newsletter is out. It's about the boring part of compounding: the long stretch where you're doing the reps and the line barely moves. Featuring James Clear's exponential growth curve, Morgan Housel, Buffett, content, fitness, and why most things work later than we want. Link below 👇

Cool thing of the day: generated this part with @OpenAI Codex + @ForgeCAD, and now 3D-printing it on my @BambulabGlobal printer (it's an obscure rowboat part FYI)

9/9 The better question: What work is slow because it genuinely needs more people, and what work is slow because the org design adds drag? That distinction is going to matter a lot. Full essay: grahammann.net/blog/ai-nativ… Sources: @brian_armstrong / @coinbase memo and @ttunguz on resiliency.

8/9 Small AI pods can be much faster. They can also be fragile. The useful question is not "can we replace 15 people with 3 people and agents?" Sometimes yes, sometimes no.

7/9 @ttunguz had the useful counterweight: "The tradeoff at the heart of AI/labor ratio decisions is not throughput. It is resiliency." A 3-person AI-heavy team may be fast. But one resignation can take a third of the human context with it.

6/9 The risk is that companies confuse cuts with redesign. You can reduce headcount, buy AI licenses, and still keep the same slow approval chains underneath. That is not an AI-native company. That is a smaller old company.