Notable verbatim quotes

All line/section references point into transcript.md. Quotes preserve transcription artifacts — do not silently correct.

On the core thesis

• Section 5 — "Our main takeaway was with AI bad things get worse and good things can get better."
• Section 2 — "adopting a gentle coding practices [sic — agentic coding] and getting most out of AI is not just as simple as, like, giving people some poor code licenses. There are changes that need to happen around and alongside that."

On prerequisites / fundamentals

• Section 6 — "if you are not doing software development well at the moment and throw a gentle coding at the problem, things are going to get worse... you're optimizing a system, you're making one part of an interconnected system go really fast, you will expose bottlenecks elsewhere."
• Section 6 — "if you don't have a platform like you can't ship code reliably and quickly and then you're just going to generate all this code out of nowhere to go with it."
• Section 6 — "If agents can't run tests to find out they've broken your software, don't be surprised when they break your software."
• Section 6 — "If the humans in your organization don't agree what good looks like then there's a very low chance that an agent is going to be able to produce code that your team is going to approve of."
• Section 6 — "the tricky part is that nobody knows where this balance point is. Like how crap do you have to be?"

On training technique

• Section 7 — "liberating structures is like a menu of different meeting formats and ways of facilitating conversations that tend to subvert power dynamics and make sure that idiots with loud voices like me don't dominate the entire conversation."
• Section 9 — "a really powerful technique is after just getting somebody to do an exercise is what did we just do?"
• Section 9 — "telling them the first half of the day no laptops. We were going to talk it out. It also made the training stick more."
• Section 10 — "You're better off having no instructions in your agent's MD then lots of them."
• Section 10 — "We don't need to train people on how to use coding agents on a happy car [path]. It's when it goes wrong."

On metrics & outcomes

• Section 11 — "we're seeing AI adoption now around 94% based on did you use AI in the last 10 PRS 94% said at least in at least three or more 60% in 10 out of 10 and then a bit less for only 6% of organization did not use it."
• Section 12 — "how do you measure a successful AI transformation? It is how many more bold things are you doing? How many experiments are you trying out?"
• Section 12 — "something that we built for eight years we were able to rewrite in three weeks teacher parity [sic — feature parity] one to one."

On SDLC reinvention

• Section 13 — "they will do like seven or eight passes of a genetic code review before then raising a PR which they will then self merge so they don't have the bottom neck of other humans reading code."
• Section 13 — "humans doing stat analysis of code is a silly idea on birthplace [sic — likely 'in the first place']."
• Section 13 — "trying to use the existing scrum Kanban methodologies is you're going to hit bottlenecks very very quickly so once you've got those practices in place then it's time to reassess your SELC [SDLC]."

On people & change

• Section 14 — (about Dominic, the converted sceptic) "he was an AI sceptic when we came to the training... it was when he had the opportunity to play with Opus 45 back in November and freedom to explore that that's when he realized that the previous beliefs weren't true."
• Section 14 — "I haven't written any code at work for four or five months... he's doing it for fun at home, you know just like you might buy furniture from Ikea but then do carpentry in your spare time."

On cautions

• Section 16 — "we are having some of our engineers saying I can't turn it off. It's 6 p.m. 7 p.m sometimes 1 a.m the dopamine in Russia [sic — rush] is real."
• Section 16 — "people are working overtime because of the increased feature pressure of you've adopted a genetic [agentic] coding you should be getting loads more stuff."

On the vision

• Section 17 — "we want to pivot from software we built for for our opcodes [opcos] to opco's build software with our support."
• Section 17 — "we see a world of more software that is more bespoke more usable solving individual people's business problems rather than kind of big slow software."