Both are right, at different stages of the same feature
AI has made building a working version of software cheap. It has not made being wrong in production cheap.
Read the note → Notes
Operating notes
One practical idea each, drawn from the papers and the live mandates behind them.
The most revealing line in Amazon's AI memo
Everyone is quoting Amazon's warning against "tokenmaxxing" this week. But the most revealing line in Amazon's new engineering memo isn't about the engineers avoiding AI.
Read the note →Diligence can spot the delivery discount. It can't fix it.
Most software owners prepping for a sale do the standard cosmetic cleanup: cap professional services on the P&L and present a "clean" SaaS revenue mix. Sophisticated buyers don't buy it.
Read the note →Different scale. Same operating principles.
Andy Jassy’s operating principles at Amazon map to the rules we use to turn around 25-to-300-500 person B2B software companies.
Read the note →Most companies didn't replace their junior engineers with AI.
Most companies didn't replace their junior engineers with AI. They just gave their senior engineers a second job.
Read the note →Every legacy SaaS company has an empty chair in the deal room.
Usually, it costs them more than a full turn of revenue on their exit multiple.
Read the note →Distributed AI governance isn't just a regulatory concept
Distributed AI governance isn't just a regulatory concept—it is the best mechanism we have for training safer, more aligned systems.
Read the note →A software business on the pitch deck. A services business in the P&L.
And the market prices it exactly that way.
Read the note →How to Get a Higher Multiple Without Growing Revenue
Delivery is not just a cost line; it is a valuation input. In a company that took on custom work to grow, it is the largest input nobody is actively managing.
Read the note →How to Scale a Tech Company Without Starting With Technology
Systems scale whatever shape is already there. Org design, people, process, systems — the strict order, arrived at by trying it backwards.
Read the note →
