Creator perspective · MoneyArik argues there will not be a world in which you can be completely hands-off with agents; the question is how much you leave to the machine versus how much you control. He recommends defining the specification, what is important to you, and what success looks like, and continuing to have people involved with agent-driven development. He also says software developers need to become architects, orchestrators, and managers of agents, and that more implicit knowledge needs to be made explicit, with people's job being to determine where to question AI output and where to poke.
endorsed
AI Engineer · Redesigning How Software Gets Built With AI Agents — Sonar & McKinsey Panel
Creator perspective · MoneyWhat you want at any given moment is a chip precisely targeting the workload you care about, in your hand today, in volume — but you'll want something else again in six months. AI workloads move on very fast — a new model comes out about every two weeks. Sanity comes from looking across models and identifying what is common to all of them.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Principle · MoneyHaving a rolling frontier of bets that you hope are deeply aligned and are ready to trigger the ramp of is a huge advantage — it gives a perennial six-month gap against competition. A rolling frontier of deeply aligned bets that you are ready to ramp at any given moment is a huge competitive advantage if you can build that machine and engine.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Report · MoneyWalter Goodwin describes Fractile's core pursuit over the past four years: building chips that run huge models much, much faster than the chips of today, scaling to models beyond the frontier and to extraordinarily long context.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Creator perspective · MoneyGoodwin describes an asymmetric game for frontier labs: going all in on a hardware bet exposes them to enormous risk, because a competitor's computational breakthrough might only work on a different chip. In a hypothetical, lab one goes all in on proprietary silicon; if a competitor discovers a chip-locked breakthrough, lab one could die in the nine months before it deploys enough of its own chip to gain a 5x computational efficiency advantage the competitor already has. Because of this asymmetric risk, the world sustains third-party frontier chip players over multiple decades rather than rolling up into the labs. Frontier labs and chip players are playing different games: labs compete in the model layer, while chip players should avoid deeply differentiated all-in bets.
hypothetical scenario, not an observed event
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Creator perspective · MoneyGoodwin explains Fractile's workload bet: hard problems with intrinsic wall-clock latency in their execution step will demand more thinking before each experiment, so Fractile aims to be the chip that accelerates the solution to hard problems in general. This bet informed the company's architectural response, alongside the bet that the industry would slip into a deployment era.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Principle · MoneyThe needed chip property is incredibly high bandwidth to memory — to load weights and state many thousands of times per second — combined with economical memory.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Principle · MoneyA chip company must span a duality: build something exceptionally good at the gamut of workloads coming down the pipe under the hardware lottery (trained on HBM-based GPUs or XPUs), while also exploring new capabilities. Building chips with fundamentally new capabilities lets you explore whether there are gravitational pulls you get to exert on the new model landscape because of those properties.
No Priors: Artificial Intelligence | Technology | Startups · Frontier Chips for Frontier AI Labs, with Walter Goodwin, Founder/CEO of Fractile
Creator perspective · MoneyMatt Diggity argues that knowledge is commoditized now — tools like cloud code can tell people how to do SEO themselves and do much of the work. He started thinking about pivoting into things AI can't do, like human relationships, noting there are still transactions made from human to human and that in-person events are something AI won't be able to touch. He is banking on those things being craved even more.
Ahrefs · He Built A $10M+ SEO Empire. It Nearly Broke Him.
Creator perspective · MoneyIf asked six years ago, the speaker would have described enterprise deployment as a technology problem, but it is actually a change management problem. Change is really hard at scale in large organizations, where every single individual has a slightly different objective function. The speaker believes change management is the number one biggest value driver for the next decade in the AI moment. Valon has spent the last six to 12 months building its change management muscle. Doing enterprise — especially large, complicated, regulated enterprise — is unlike what you would imagine thinking about it in the abstract.
The a16z Show · How Valon Rebuilt a $13 Trillion Industry From Scratch
Principle · MoneyDor Sasson argues that selling AI at scale requires building banking-like systems — reservation, reconciliation, ledgers — and that AI companies should be aware of all the required banking-like traits from the start rather than treating each case as unique.
AI Engineer · Every AI Company Is Accidentally Building a Bank — Dor Sasson, Stigg
Creator perspective · MoneyIn the architecture most teams ship today, checks and settlements happen after inference and after usage is logged — which is too late. This produces spikes, overspends, overages, sticker shock, bad user experiences, and potentially a margin problem for vendors who didn't allocate resources for unconstrained usage.
Speaker's characterization of common practice
AI Engineer · Every AI Company Is Accidentally Building a Bank — Dor Sasson, Stigg
Creator perspective · MoneyDor Sasson argues that the AI pricing emergencies are not merely commercial problems but infrastructure emergencies: Anthropic had no effective way to delineate API users from subscription-plan users, and the common failure across companies is that entitlement checks happen after the invoice, not before.
Speaker's interpretation of the events
AI Engineer · Every AI Company Is Accidentally Building a Bank — Dor Sasson, Stigg
Creator perspective · MindJoe Liemandt explains that when kids push through hard things — by whatever motivation it takes — they come to see themselves as capable, and that new self-perception creates intrinsic motivation going forward; showing kids they're capable and breaking their blocks is the unlock. He illustrates this with his daughter, who after hitting top 1% told her mom the achievement changed how she sees herself.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Creator perspective · MindJoe Liemandt describes how kids have blocks ('I can't do that, I'm not that kind of person') and argues that a job of schools and educators is to crush those blocks and show kids the belief isn't true; the best teachers change lives by seeing a block and convincing students they can do more. At his school, guides work daily on finding and removing each kid's block.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Creator perspective · MindLiemandt argues that if America applied the sports/athletic ethos to academics, it would fix every academic problem in the country, citing that parents want their kids on the Westlake High School football team because it will be the hardest and most fun thing they ever do and they'll remember it the rest of their life, and that education can be made the same way as sports — hard, fun, and memorable.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Creator perspective · MindLiemandt argues that what changed in athletics is not genetics but the training and instruction: society knows how to train people to reach their physical frontier. Athletes get coaches and structured, step-by-step instruction rather than being told to 'do your best' or figure it out alone. He proposes covering the athletic side first, then applying the same logic to academics.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Principle · MindJoe Liemandt states as a principle that doing hard things in one domain you love builds grit that transfers to hard things in other domains.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Creator perspective · MindLiemandt argues that school socialization is a real part of school — it's much more than sitting and watching a YouTube video — and that watching YouTube videos provides materials but not motivation or structure, which schools provide a lot of. He adds that motivation is 90% of learning these days.
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt
Principle · MindLiemandt argues schools need to create builders — that's what kids and adults should be doing; you can't just be thinking or sitting in class all day. When kids get to build — physically in sports and climbing, building drones and robots, building a business, writing books, making videos — that's what engages them and what they love, and he says we need to bring that back. He contrasts this with the current educational system, which he says generates consumers rather than builders, and notes parents are limited because they went through that same terrible system, which did not develop kids to their utmost potential.
endorsed argument by the guest
Andrew Huberman · How to Accelerate Learning & Improve Education | Joe Liemandt