146: Weak signals in AI – or it grew really tall and fell over

146: Weak signals in AI – or it grew really tall and fell over

More apps are shipping to the app stores, but the same number of apps are getting downloaded. In this one we're pushing a pram round Bournemouth with a one-month-old and making sense of the current AI moment — from the slop machine that online advertising has become to why the big frontier model bet may have misfired, and what a walking creature that just falls over can teach us about software development. We feel our way through some stuff: * "Trend slop" ... how every frontier AI model tends to give you the same strategic advice regardless of context (and what does change its answer — infuriatingly) * Why running a team of AI agents is starting to sound suspiciously like managing that guy from Memento * The harness that matters more than the model — and why the frontier model arms race may have already peaked * "Boring tiny tools": the overlooked category of software that now makes sense to build (until the pricing changed) * What an evolutionary algorithm that learned to just fall over reveals about agentic engineering realities * Why growing software might be better than thinking we can build it Jus' tryna think clearly about AI and block out all the hype and panic. Find out more about us and our work at crownandreach.com ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

145: What Xerox PARC actually got wrong

Xerox PARC famously invented the mouse, the graphical user interface, and the printer — then watched Steve Jobs walk off with all of it. The usual lesson: silly management didn't recognise what they had. The actual lesson is more useful ... and more uncomfortable for innovator types. Sit in your beanbag so you can't storm off. In this one, we use Xerox PARC as a case study in what happens when an uncertainty bubble has a membrane that's too thick and too long-lived. Then we tell the story of a client engagement where we built one that had the right kind of membrane. What does it take to run the scary process of innovation under the radar, while keeping stakeholders just comfortable enough not to kill it? * Why the PARC director's real mistake wasn't being too far ahead, it was the trade fair * "Duringmath": why the groundwork has to happen during the bubble, not after * Nemawashi, information radiators, and the meetings before the meeting * The MAYA principle: Most Advanced Yet Acceptable — and what happens when you violate it catastrophically * How a challenge month became an uncertainty bubble that nobody could see was improvised and revised daily * The barbell strategy for uncertainty: work the certain end and the wildly uncertain end; leave the middle for later * Why successful projects end up nothing like the plan — and why that's the point * Pretotyping: build nothing if possible, cobble something together if necessary, build the real thing last This one's for anyone who's tried to run an experiment inside a company that wasn't ready for it, or tried and failed to innovate by following official processes, or wondered why the brilliant innovator always seems to end up persona non grata. References * Alberto Savoia, author of The Right It and coiner of "pretotyping" * Uncertainty Bubbles, concept by Crown & Reach * MAYA principle — Most Advanced Yet Acceptable, by Raymond Loewy. * Nemawashi – the Japanese principle of preparing the ground; pre-meeting meetings * Duringmath – our new coinage from this episode :) * 4U framework: Unpack, Undergo, Unfold the Uncertainty, by Crown & Reach * Barbell strategy, referenced in this episode, should be attributed to Nassim Nicholas Taleb * Deliberate strategy vs emergent strategy (attributed to "a Japanese CFO" quoted to by JP Castlin Find out more about us and our work at crownandreach.com ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

144: The domino rally delusion

What do you do if Strategy Doesn't Exist? Tom and Corissa have built a strategy consultancy and they're here to tell you that strategy doesn't exist. At least, not in the way everyone sells it — and probably not in the way you've been trying to do it. This one was sparked by a recent episode from friends-of-Tentacles Kyle and Jen on Notes from the Swamp. We start with domino rallies and end up somewhere genuinely uncomfortable: what if the thing you're planning isn't real, the thing that worked was partly luck, and everyone who sells strategy is either in on the joke or still under the spell? * Why "if you just set the dominoes up right" is a fantasy ... and why we all keep buying it anyway * Deliberate vs emergent strategy is a taxonomy that sounds useful ... until Corissa neatly dismantles it * The project pattern that's almost too perfect to be true: everything that succeeded ended up nothing like anyone could have expected at the start; everything that failed stuck to the vision * Henry Mintzberg, retrospective coherence, and the stories we love to tell ourselves about how we got somewhere good * Can your immune system be strategic? (Tom thinks yes. Corissa is not convinced and does not enjoy the question) * Why Dave Snowden can sell the messy, hard version ... and why most people can't (at least not yet) * The Trojan horse question: do you know what's inside your horse? (Is it another, smaller horse?) For anyone who's ever suspected that "strategy" is a collective polite fiction and is curious about what to do with that suspicion. REFERENCES * Kyle Godbey and Jen Briselli's Field Notes from the Swamp podcast https://www.youtube.com/watch?v=TELdIQUd8nM * Henry Mintzberg — strategist, referenced for his empirical work on how strategy actually plays out * JP Castlin's Strategy in Praxis substack; referenced for a four-part taxonomy of strategy types https://strategyinpraxis.substack.com/ * Michael Porter and classic strategy doctrine. Tom accidentally said he came up with the seven S's framework, that's actually McKinsey; Porter is associated with Five Forces and competitive positioning. * Dave Snowden and the Cynefin Company, referenced for practitioner-level rigour and the paradox of "meet people where they are" vs "don't dumb it down" * Jacinda Ardern, New Zealand Prime Minister during COVID; referenced as an example of deliberate strategic choice * Taylor Pearson's expansion of John Salvatier's point that "reality has a surprising amount of detail" https://taylorpearson.me/interestingtimes/reality-has-a-surprising-amount-of-detail/ * The OODA loop / John Boyd https://thecynefin.co/a-lamb-in-wolves-clothing/ * Crossing the Chasm (Geoffrey Moore), referenced implicitly via "early majority / late laggard" diffusion diagram * SWOT analysis, referenced as a "late laggard" strategy tool * Deliberate vs emergent strategy taxonomy (via JP Castlin's Mintzberg review series) * "The three Is" https://thecynefin.co/anthro-complexity-1-3 Find out more about us and our work at crownandreach.com ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

143: The breadmaker trick

Bread makers don't knead dough the way a human baker does. They get to the same result via a completely different route — and that turns out to be a pretty good map for what's actually happening with AI right now. We start with brioche, end up in a crisis of professional tacit knowledge, and find a surprisingly useful frame for thinking about what machines can and can't do — and what that costs us in the long run. * Why "harness the model" is the real skill — and what that actually looks like in practice * The 50/30/10 split: surprisingly good, fine, and catastrophically bad — in the same tool (and no the maths doesn't add up) * Why test-driven development went from "boring best practice nobody does" to "the only way this works at all" * The tacit knowledge cliff: what happens in 20 years when there are no senior lawyers who did the grunt work * Explicit → explicable → tacit → relational: a spectrum that explains where AI taps out * Why scarcity and skin in the game are the two things a language model structurally cannot fake * Artisanal lawyers, peak athletes, and the industrial revolution: why commodification always leaves a pocket for the handmade For anyone trying to think clearly about AI without falling for either the hype or the backlash. References * Luca Dellanna's piece on what makes humans different from AI * Nonaka & Takeuchi (1995) – the bread maker as an example for how knowledge is encoded in processes and organisations https://www.scribd.com/document/258487259/Nonaka-I-and-Takeuchi-H-1995 * (Found after we recorded) a critique of Nonaka & Takeuchi's work on bread making machines - bread maker not as incorporation of tacit knowledge, but as fitting a social prosthesis into a rearranged world: https://www.researchgate.net/publication/215439406_The_Bread-Making_Machine_Tacit_Knowledge_and_Two_Types_of_Action Find out more about us and our work at crownandreach.com ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.

142: Sit with it

When a hiring manager asks an unanswerable question, what if that's the whole point? In this one, Corissa brings along a Reddit thread that stopped her in her tracks. A candidate who discovered, mid-interview, that the interviewer had never once rejected someone based on their actual answer. What they were watching for was whether you'd sit with the discomfort of having no right option, or immediately reach for a safe response ... one where you try to please the interviewer. That question spirals us into the Kobayashi Maru, the art of giving feedback (and filtering feedback), why founders hold back from the conversations that would actually tell them the truth, and what it means to stop preparing the right answers and start trusting your own judgement. * The Star Trek test designed to be unwinnable — and the eejit who won anyway * Why the interviewer never rejects based on the answer itself * The golden rule for making sense of feedback that's almost always true * Why founders often make every move except the one that would actually test the idea * Why when you hear "other people would ..." it should set off alarm bells * The Red Queen effect in interviews: why every clever question eventually gets gameable * How to prepare for interviews by stopping preparing for interviews This one's for anyone who's noticed that the world keeps getting less predictable, and that "sitting with discomfort" is somehow both obvious advice and surprisingly hard to take. References * The Iron Triangle (good / cheap / fast ... pick two) * Mike Haber's Inverted Iron Triangle (bad / late / over-budget ... you can have all three!) * Neil Gaiman's feedback rule (attributed, then immediately un-attributed, could be apocryphal) * Star Trek's Kobayashi Maru scenario * The Red Queen Effect Find out more about us and our work at crownandreach.com ---------------------------------------- Hosted on Acast. See acast.com/privacy [https://acast.com/privacy] for more information.