{"version":"1.0","type":"rich","provider_name":"Acast","provider_url":"https://acast.com","height":250,"width":700,"html":"<iframe src=\"https://embed.acast.com/$/68357ec21b846c88bdcd7480/6a3e728889bd872840ca6ef6?\" frameBorder=\"0\" width=\"700\" height=\"250\"></iframe>","title":"146: Weak signals in AI – or it grew really tall and fell over","description":"<p>More apps are shipping to the app stores, but the same number of apps are getting downloaded. </p><p><br></p><p>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.</p><p><br></p><p>We feel our way through some stuff:</p><p><br></p><p>* \"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)</p><p>* Why running a team of AI agents is starting to sound suspiciously like managing that guy from Memento</p><p>* The harness that matters more than the model — and why the frontier model arms race may have already peaked</p><p>* \"Boring tiny tools\": the overlooked category of software that now makes sense to build (until the pricing changed)</p><p>* What an evolutionary algorithm that learned to just fall over reveals about agentic engineering realities</p><p>* Why growing software might be better than thinking we can build it</p><p><br></p><p>Jus' tryna think clearly about AI and block out all the hype and panic.</p>","author_name":"Tom Kerwin"}