Creator perspectiveThe biggest problem with geostationary satellite internet is distance: a request from a plane to a geostationary satellite travels ~140,000 km per ping, taking almost 500 ms at the speed of light, plus ground latency. 2000s airplane internet (and much of today's) is slow because it operates with a necessary latency of at least 450 ms, sometimes spiking over 3,000 ms. There is currently no workaround for the speed of light limiting satellite internet latency.
creator's explanation
Slidebean · The fatal flaw with satellite 5G
ReportThe big innovation with Starlink and the upcoming Amazon Leo is that the satellites are no longer far out in space — they're in low Earth orbit. Low Earth orbit is 330 kilometers instead of 35,000 kilometers, dropping latency from 500 milliseconds to just 2 milliseconds round trip.
none stated
Slidebean · The fatal flaw with satellite 5G
Principle · MoneyA common anti-pattern: people post their first video, decide it's horrible, and stop posting — but the important part is to continue doing it. Founders should spend even 10 minutes a day stepping back to ask: what's something I did today that made me smile or was interesting or felt new? The daily launch routine: talk about something interesting from your day, show it, click post, then close it and get back to work — that's all you have to do.
Y Combinator · This Is How You Should Actually Make Launch Videos
Principle · MoneyFarza's video formula is always the same: this is who I am, this is what's going on, this is what I'm working on, let me show you. Even a simple video — 'hi, this is what I'm working on, check this out' — then turning the camera to literally show your computer can work, though the effect may not be as big.
Y Combinator · This Is How You Should Actually Make Launch Videos
Creator perspective · MoneyJassy explains Amazon structures capital investment in two stages: data centers are committed roughly two years in advance (30-year useful life, one-time investment), while servers and networking gear are bought only a few months in advance so they can be cancelled if demand doesn't materialize. If capacity turns out unneeded, the flexible later-stage commitment protects against overbuilding.
described as Amazon's actual practice
The a16z Show · Amazon CEO Andy Jassy on the $220 Billion AI Buildout
Creator perspective · MoneyWhen debugging why an agent did not call a tool, first break down the entire context available to the agent and check whether the tool was even available. If the tool was not available, stop there and fix that particular scenario before investigating further down the pipeline. Continue along the pipeline of available information, checking whether the tool was called and what values were returned.
AI Engineer · Don't Trust Your AI Agents. Test Them Like a Scientist — Stephanie Jarmak, Omni Analytics
Creator perspective · MoneyThink about where the human-review bottleneck might be, how to change what the alarm rates tend to be, and how consequential those alarms will be for downstream effects. An LLM-as-a-judge system with a 1% false alarm rate would generate 10,000 human reviews if all alarms went to humans, which may exceed your organization's domain-expert capacity.
AI Engineer · Don't Trust Your AI Agents. Test Them Like a Scientist — Stephanie Jarmak, Omni Analytics
Creator perspective · MoneyAn agent can appear to receive information while actually not having it, purely because it did not call the tool that would provide that information. In one case, the agent received certain information and then trusted that that information wasn't there, purely because it didn't use the tool.
AI Engineer · Don't Trust Your AI Agents. Test Them Like a Scientist — Stephanie Jarmak, Omni Analytics
Principle · MoneyKeep native scroll behavior because when the page keeps scrolling after you stop touching the mouse or trackpad it feels unnatural and messes things up. Reference top Fortune 500 company sites with millions of page views as the benchmark: they do not scroll jack.
Y Combinator · How To Stand Out In A World Of AI Designs
Principle · MoneyWhen using tabs on a landing page, you have to assume people will not see anything behind the other tabs. On some sites the tab timer starts before the user sees it, so everyone is actually seeing only snippets. Most users are not looking at a site for the first time with deep scrutiny; they won't invest the effort to decode a complex interface. Showing a demo on a landing page is cool but can have too much going on and too much text to mentally orient around. The demo plays too fast for the viewer to understand what is being typed and what the product is doing. Consider replacing a full-screen screen-recorded demo video with a simpler chat-inbox-style animation, or get an agency to produce a really cool animation for the landing page instead of a full-screen video.
Y Combinator · How To Stand Out In A World Of AI Designs
Creator perspective · MoneyAnekha Sokhal describes breaking a market down into its component data types rather than treating it as one blob: structured data (futures time series plus technical indicators, and fundamentals time series covering demand and supply), unstructured data (news, X/Twitter, podcast audio), and the user's private data (trusted analysts, subscriptions, internal data).
Described as her own approach with moderate confidence; the atom carries a possible_numeric_error provenance flag.
Latent Space · You are the Team — Anekha Sokhal, Moshi
Principle · FuelHuberman emphasizes that the distinction between stimulant-based and neuromodulator-based supplements is very important and often overlooked: caffeine and alpha-yohimbine mostly increase alertness but don't tighten focus, while alpha-GPC and L-tyrosine provide less alertness elevation but more focus tightening. The two approaches are simply different, neither better nor worse.
Andrew Huberman · Essentials: Developing a Rational Approach to Supplementation for Health & Performance
Creator perspective · MoneyEvan Conrad predicts that without compute markets, AI will consolidate into a TSMC-like structure: one or a few buyers that are the clouds, a few model labs, and no other startups.
asserted
AI Engineer · Why AI Startups Will Disappear Without Compute Markets — Evan Conrad, San Francisco Compute
Report · MoneyThe conversation reports that developing a drug takes 15 years and $1.5 billion: about $15 million for preclinical work such as mouse studies and the molecule versus $1.4 billion plus for clinical trials, and roughly two to three years preclinical versus another 12 years for clinical trials. Drugs take 15 years to market partly because successively more expensive trials with different safety considerations culminate in large-scale human trials.
No Priors: Artificial Intelligence | Technology | Startups · The Changing Physics of Cybersecurity and Drug Development with OpenAI President Greg Brockman
Report · MoneyBrockman asserts it is 'impossible to deny' that AI can now generate genuinely new knowledge — resolving century-old math problems — though it has not yet produced whole new grand theories. OpenAI has published resolutions to almost 400 important open math problems, with domain experts finding surprising applications of ideas in the AI's work.
partial
No Priors: Artificial Intelligence | Technology | Startups · The Changing Physics of Cybersecurity and Drug Development with OpenAI President Greg Brockman
Creator perspective · MoneyAn open-source tool called REA (Reverse Engineer Anything) harnesses an AI model: point it at any piece of hardware and it will reverse engineer it and generate new source code. You could point it at a malfunctioning thermostat, have the AI write and install new code with a new interface. Basically any form of hardware can now be taken over and reprogrammed with this stuff — one idea is an AI bot that takes over all the hardware in your house and runs home automation the way you want, which the speaker says friends are literally doing right now. Anyone who understands software may automatically be able to do things in hardware that were historically difficult or inconceivable.
moderate
The a16z Show · The Frontier Ideas a16z Wants to Fund | Marc Andreessen, Ben Horowitz & Chris Dixon
Principle · BodyThe host calls the jump rope the bridge to getting twitchy, with the lowest barrier of entry and the best warmup. For a beginner it's arguably the exercise you could use for six months; he recommends spending the first three months dialing in jumping rope if you're developing a bigger vertical jump, doing five sets of about 60 seconds to prep for a jump session, and after basic jump rope starting to play around with one leg at a time.
Garage Strength · I Trained 1000 Elite Athletes. This Is How You Jump Higher.
Principle · BodyTo increase your jumps, the host recommends two minutes of rest between sets. The goal of the training is to get you twitchy and explosive — not to do a CrossFit workout. During rest periods, regroup and think about the technique and ground reaction as quickly as possible.
Garage Strength · I Trained 1000 Elite Athletes. This Is How You Jump Higher.
Principle · MoneySarah Chang says she and Kova generally follow six stages — ideating, scripting, recording, editing, repurposing, and publishing — and the workshop is structured around them.
AI Engineer · The AI Content System Behind 200M Views — Sarah Chieng, Cerebras & Cova Lee, OpenAI
Creator perspective · MoneyBouchard describes the failure modes of sloppy AI text: it is too general and shares nothing specific, repeats itself, hallucinates so it can't be fully trusted, and — worst — has the wrong tone, not talking like you. Overused LLM words like 'delve' are a marker.
AI Engineer · Why AI Writing Still Reads Like Slop — Louis-François Bouchard, Towards AI