35,595 results · Most relevant
Liemandt says an AI tutor can assess exactly what a student knows and doesn't know, give a personalized lesson, and keep the student in a loop to mastery before moving them on. He calls individualized AI lesson plans to catch kids up to grade level the number one use of AI in education — not chatbots — and the single best thing to fix education in America.
the guest's endorsed recommendation
Keep up with the model treadmill with the latest models. As new models come out with better reasoning capabilities, continuously evaluate which model is suitable for the task and update the right model for your use case. The orchestration engine for moving across models in a workflow needs to keep pace with how models are evolving and marry the best model for the best task — e.g., Gemini for image-specific reasoning, OpenAI for deterministic steps, Claude for open-ended investigations.
Dr. Masud Husain advises the students he teaches — who are probably not in the low baseline state — to think twice about taking a stimulant to get a better grade, because the evidence suggests they might actually get worse from doing it.
High-performing students thinking of taking stimulants for better grades risk making their performance worse, since those already at high dopamine baseline get no benefit. Similarly, nicotine users escalating their dose are likely just maintaining their normal baseline rather than gaining cognitive benefit.
Consequential creator advice requires independent verification before Today can promote it.
Post raw fly-on-the-wall footage of yourself actively coaching athletes, with a videographer clipping the best moments. Lean into being an active coach as a content differentiator, because it's a strength above 99% of people on social media. A 10-minute unstructured clip of coaching heavy lifting (like Dane coaching Sam's 440 pull) delivers 2-3 tasty morsels of applicable value while entertaining viewers. This content should not be structured or a super hype MrBeast-style edit.
uncertain
For an established creator with a strong audience, posting 80% high-TAM content would eventually cause the core audience to switch off, so Eugene recommends scaling back to roughly 20% high TAM. He specifically advises the host to experiment with bumping high-TAM volume up in the short term since he hasn't done much of it.
Eugene Teo recommends picking high-TAM topics that feel adjacent to your niche but are more general to the population, like 'red meat gives you cancer'.
Consequential creator advice requires independent verification before Today can promote it.
If just starting out, Eugene says spam the high TAM to get people knowing you — around 80% high-TAM content — and practice the skill of tightening a script for a general audience, speaking in simple, relatable language that isn't coach speak (not dumb, just simple). He himself had to unlearn fancy coach-speak words or calibrate them to the right person.
Ship a change to legacy users only after you can confidently answer yes to whether the change was explained, workflow breaks are necessary, and users have time or choice to adjust.
Pollan recommends running a caffeine fast to understand your relationship with caffeine — get it completely out of your system. He says to give it a try: you learn a lot. The host asks what an experiment like quitting caffeine for an extended period could do for someone, perhaps allowing caffeine to be an even better experience afterward.
Huberman recommends a 30-second to three-minute visual anchoring period at the beginning of a 90-minute work or physical work bout: set a timer, do the anchoring (covertly if you prefer), then move to the mental or physical work you're trying to do. The practice ramps up neural activity within the circuits that create focus and concentration, recruiting the prefrontal cortex and frontal eye fields.
Hormozi advised Caleb to chase the repairs, not just the big roof replacements, while raising prices and increasing volume.
Hormozi advised Joey to pursue B2B events and sponsored booths, which put him in front of hundreds of ideal customers at once instead of chasing individual accounts one at a time.
Hormozi advised Joey to restructure the sales team so they set appointments while the owner closes, add the events-per-year question, deploy three VSLs, do same-day quotes, and run a daily LinkedIn outreach cadence.
Thariq recommends Claude Tag for tasks requiring shared context and permissions — legal reviews, incident management — because incidents are inherently multiplayer and permissions are handled in Slack; he describes running a channel per project so legal can query the agent directly and he doesn't need to be in the loop.
A locally coherent method for bench dips: (1) externally rotate the shoulders by pointing the fingers out to the sides or even behind you; (2) open up the chest; (3) press up and back from the bottom to get a good triceps contraction without the shoulder negatives. With externally rotated hands, the shoulder avoids forward positioning at the bottom, creating a friendlier position for the shoulder joint.
Andrew Bialecki recommends defining levels of autonomy for all employees — not just engineers — and setting a deadline (end of June) for everyone to reach L3, meaning constantly running agent sessions or a team of agents. His rationale: if you're not constantly running agents, you won't be skilled to survive in the next era.
The creator challenges viewers to pick one thing they would do with just 10% more confidence — ideally something that improves self-esteem no matter how it goes — with examples like cracking a joke at checkout, starting a conversation with a random person in the office, or telling someone a truth they have been avoiding. The instruction: do not judge the outcome, put in the first rep, because taking the opportunity already means you have succeeded.
To present more confidently, the creator recommends opening up your body language and showing your vitals, slowing down your movements, and telling your own embarrassing stories without shame. He flags a trap: these techniques are not for showing others how confident you are.
Gabriel Chua warns that a software pipeline without a testing-and-approval layer lets PRs land without human validation. He recommends adding layers where a test version is built, sent to someone for testing, gets a thumbs up, and that feedback is fed back to the PR so the reviewer can see it holistically.
Choose sources → to get a library of your own and a daily brief.