Principle · BodyDuring acceleration the trunk should be fixed at ground contact and not keep bending forward or collapse. Trunk collapse is visible only in qualitative high-speed video analysis, not in any of the objective numbers. When analyzing sprint video, look at the overall position of the segments just before contact and during the first millisecond of contact, then the overall contact phase and the maximum collapse and trunk lean angles the player has to support.
Garage Strength · Why Most Athletes Train Speed Wrong w/ Sprint Scientist Dr. Jean-Benoit Morin
Creator perspective · BodyDr. Morin explains that resisted sprinting always comes with longer contact phases, so you need to add complementary high-speed training or high-speed plyometrics to work on the contact time part of the equation. If you do only resisted sprinting, you may improve force output but within a contact duration that is still not short enough to be specific to high-speed sprinting. He also notes that after a hard session of resisted sprinting, athletes will complain about DOMS or pain in the calves, indicating a significant overload in the calf muscles.
Garage Strength · Why Most Athletes Train Speed Wrong w/ Sprint Scientist Dr. Jean-Benoit Morin
Creator perspective · MoneyConnecting agents to messaging platforms like Slack, Teams, Discord, WhatsApp, or Telegram requires multi-step manual setup (Telegram 5, Discord 7, Slack 8, WhatsApp 11 steps) and heavy documentation, and yields only an agent talking to a person — the platforms are built for humans and actively block bot-to-bot connections, leaving agents in 'digital solitary confinement'.
AI Engineer · Why Your AI Agents Can't Talk to Each Other (Yet) — Vlad Luzin, BAND
Principle · MoneyA failed agent run is the richest signal available because it exposes missing guardrails, missing libraries, and unconsidered edge cases; treat failed runs as primary input for improving the skill rather than discarding the trace.
AI Engineer · From 36% to 100%: How Self-Improving Agents Write Their Own Skills — Rafal Wilinski, Runlayer
Principle · MoneyWhen evaluating AI pen testing vendors or in-house solutions, the most important thing to remember is that context is everything — LLMs are as good as the context you feed them. Feed the solution with data from all previous scanners (static, dynamic, all engines) rather than running it standalone; take context from many different machines and environments and combine them to guide the LLMs. When models know where to look, results are significantly better — and richer data makes the LLM operation better and cheaper. One solution is not enough: combining the data from multiple tools is what's important.
AI Engineer · AI Hackers Are Faster Than Your Pen Test — Eli Cohen, Snyk
Principle · MindYou have to make trades; you get to pick your regrets, or they will be chosen for you. You cannot trade everything in, so you get to pick the things you say were less important — otherwise life chooses for you.
Stated as the speaker's core advice.
Alex Hormozi · It took me 37 years to realize what I'll tell you in 18 minutes.
Principle · MindYou can be a master of your own life to a degree if you can say: I'm willing to give up these other things for the things I care about most. The meta of prioritization is choosing the game you want to play.
Stated as the speaker's hope for the audience.
Alex Hormozi · It took me 37 years to realize what I'll tell you in 18 minutes.
ReportPhilipp Schmid explains that with managed agents, a single API call starts a remote sandbox that loads all files and runs the entire agent loop server-side, so developers don't need to manage function calling on their client or any infrastructure.
AI Engineer · Why AI Agents Should Have Their Own Sandbox — Philipp Schmid, Google DeepMind
Report · MoneyMichael Lee describes the three criteria Sequence uses to evaluate investment targets: scale of the market, whether the incumbent holds all the advantages, and whether the overlap between the organization's work and what AI can do today allows building something special.
No Priors: Artificial Intelligence | Technology | Startups · Re-Founding Incumbents for the AI Era with Sequence Holdings Co-Founder and CEO Michael Lee
Creator perspective · MoneyThe next evolution is giving non-technical users an interface where they can experiment — tweak a system prompt, change the underlying model, iterate on parameter values — not just log results. But the primary gap remains: test cases are still a curated set from offline evals, so once shipped to production you are operating blind — no idea what steps your agent is taking or when failures happen in the live environment.
AI Engineer · Why Building an Eval Platform Is Harder Than It Looks — Braintrust
Creator perspective · MoneyDru Knox describes the improvement loop structure: the inner loop covers the coding agent's work on a PR before it is put up (fast, cheap, iterative loops the agent runs all the time); the outer loop runs after the PR is up, where you put more expensive, exhaustive checks that encode what a human would otherwise verify (e.g., agentic QA or mutation testing to check test suite quality, run once when the PR goes up, then the agent iterates on results); the meta loop feeds back into the inner and outer loops to ensure the same mistake is not made again. He notes this is where you'll end up spending most of your time, and that you drive up AI-native AMI by investing in the meta loop once you have infrastructure for each loop.
AI Engineer · Harness Engineering: How to Build a Software Factory — Dru Knox, Tessl
Principle · MoneyMarc Klingen recommends maintaining the boundary of what you optimize against and using AI mostly for proposing improvements to the agent implementation. He warns not to overfit the application to a random quirk found in production that doesn't matter for what the agent should actually be doing, and to be in the loop to monitor the failure definition because agents find all sorts of failure patterns and some don't really matter that much.
AI Engineer · The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse
Principle · MoneyMarc Klingen warns that benchmarking on datasets that are not in sync with production means the data is inaccurate, and that monitoring production data without benchmarking offline leaves you without validation. Teams should continuously align agent datasets with what actual users are doing — for a customer support agent, the dataset should be common support questions.
AI Engineer · The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse
Principle · MoneyMarc Klingen recommends staying in the loop at the most important steps of setting direction — goals, datasets, evaluators, and failure definitions — while pulling yourself out of manual and tedious work on the lower levels of the loop. He warns that giving eval criteria and dataset changes completely out of hand risks creating more slop, and that fully automating the loop without being in it risks producing a lot of tokens that don't really make sense. Teams are usually still involved in reviewing AI-proposed changes to eval criteria and datasets because they create the boundary for other agents to auto-improve against.
AI Engineer · The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse
Principle · MoneyMarc Klingen recommends designing your application so it automatically feeds in interesting signal — for example customers swearing at a support agent, customers saying 'this was not correct', or an internal agent accepting or changing proposed messages — so the agent loop has an in-stream of signal to act on. The high-level loop should have signal coming from you, your team, or external sources, while lower-level loops can be automated.
AI Engineer · The Self-Improving OSS Agent Stack — Marc Klingen, Langfuse
ReportFigma's agent can be taught a team's design system rules and unwritten conventions by adding guidelines to the team's Figma library. Guidelines are built right into the library and teach the agent what the team already knows, so the agent uses components, tokens, and conventions without them being spelled out in every prompt.
Figma · Teach Figma's agent your design system | Add guidelines to your Figma library
Report · MindDr. Masud Husain reports that healthy aging people without dementia or neurological diagnosis who score highly on self-reported questionnaires for apathy have twice the risk of developing Alzheimer's disease than people who don't. Alzheimer's disease can lead to behavioral changes long before any sign of cognitive impairment develops, and one of those behavioral changes is apathy — loss of motivation. The pathology laid down in the Alzheimer's brain can begin more than 10 years, perhaps more than 15 years, before somebody presents to a doctor with cognitive complaints, creating a huge window of maybe 10 to 15 years that could be used for early detection and intervention.
Consequential creator advice requires independent verification before Today can promote it.
Andrew Huberman · How to Improve Motivation & Overcome Procrastination | Dr. Masud Husain
Principle · MindHusain's practical framework: to change motivation, either make the effort less (e.g., break the task down) or change the incentive to something meaningful and rewarding for that person. Framing the task as novel and salient can itself feel rewarding and lower the activation barrier, and this simple principle may be enough for someone to figure out how to motivate themselves.
endorsed
Andrew Huberman · How to Improve Motivation & Overcome Procrastination | Dr. Masud Husain
Principle · BodyYou need to be in a state of low fatigue for speed work, because training sprints hard while tired can actually make you slower. Feel fresh before speed day so your legs get proper exposure to moving at a high rate. Sprinting as hard as possible as often as possible is an anti-pattern.
Garage Strength · I Had 10 Athletes STOP Lifting Heavy. They Got STRONGER.
Principle · MoneyEugene Teo advises identifying the pain point: the emotional reason the person should stick around to watch. The emotional core can be a visceral, painful thing or a debt the viewer wants to complete and fulfill.
Garage Strength · The New Rules of Social Media Fitness (2027) w/ Eugene Teo