ChatGPT & LLMs

A working architect's view of the 2026 LLM stack — the model tiers, inference providers, orchestration layers, evaluation tooling, and the trade-offs that actually matter when shipping production AI. Everything you wish someone had laid out before you started.

Which model to reach for, by task type. GPT, Claude, Gemini, the reasoning models, and the open-weights options — sorted by what they are actually best at, with simple decision rules.

Search and AI assistants are not interchangeable. A practical guide to which tool fits which question — with side-by-side examples and the cases where you should use both.

The persistent beliefs about AI that keep otherwise sharp adults from even trying it. Each one addressed honestly — with what's true, what's exaggerated, and what to do about it.

Modern AI can read photos, charts, screenshots, and handwriting almost as easily as text. A practical guide to what works, what doesn't, and the thirty-second privacy checklist before you upload anything.

Talking to AI feels strange for about ninety seconds, then it becomes the most natural interface there is. A practical guide to voice mode — what it is great at, what it is bad at, and how to actually use it.

AI does not think the way you do. It generates likely continuations from context. A plain-English mental model that, once it clicks, makes every prompting tip in the world easier to understand.

AI does not lie the way a person does. It can generate fluent false specifics because plausibility and verification are different jobs. Here is what is going on, why it happens, and how to avoid being burned.

Four mainstream AI assistants, picked apart in plain English. A non-technical guide to choosing your first AI based on what you already use and what you actually want from it.

A jargon-free comparison of free ChatGPT, ChatGPT Plus, and ChatGPT Pro — what changes when you upgrade and how to tell if you actually need to.

A guided sixty-minute tour of ChatGPT — from signing up to your first real win. Every button, every setting that matters, in order.

AI in 2026 means one specific thing for most people — and it is not what the movies told you. A plain-English explanation of what is really going on, what it does well, and what it does not.

Y Combinator. Karpathy's AI Startup School keynote frames LLMs as a new kind of computer — utility, fab, and OS rolled together — and argues for "partial autonomy" products with a human-controlled leash. It is the cleanest articulation of the stack-level mental model the article assumes: that you are picking inference vendors and tooling for a programmable substrate, not a chatbot.

Andrej Karpathy. This is the clearest end-to-end explanation on YouTube of what an LLM actually is — pretraining, tokenization, SFT, RLHF, reasoning RL, tool use, hallucinations — at the level of detail an engineer needs to reason about model trade-offs. Watch it once and the "GPT-class vs. open-weights vs. reasoning model" decisions in the article stop feeling like brand choices and start feeling like training-recipe choices.

AI Explained. Older than the article (March 2024), but the methodology is what's useful: a single careful reviewer running the same hard tasks — OCR, theory of mind, instruction following, math — through three frontier models side by side and showing exactly where each one cracks. The model names are dated, the framework for comparing models is not.

Andrej Karpathy. Karpathy spends explicit chapters on "Be aware of the model you're using, pricing tiers" and "Thinking models and when to use them," then keeps switching between ChatGPT, Claude, Gemini, Grok, and Perplexity throughout the rest of the walkthrough. It is the closest thing to watching the article's cheat sheet applied live by someone with strong opinions about when each tier earns its keep.

Jeff Su. Jeff opens with the spectrum the article hinges on — ChatGPT and Gemini at the "creative" end, Perplexity and Google Search at the "accurate, real-time, sourced" end — and then shows you the third option in action. The Olympics-medals comparison in the middle is the cleanest "this is when you ask a chatbot, this is when you Google" demo we've seen. After watching, you'll have a third tool in your search routine and a working rule for picking between the three.

Acerola. A developer-essayist works through the "AI is just slop / AI is magical / AI is theft" trio of myths with patience and code on the screen. If the article's myth #4 or #7 ("it's just plagiarism," "it's just plagiarism but for art") nagged at you, this is the video that earns the right to disagree with you.

Cleo Abram. Cleo Abram walks through both the "it will kill us all" and "more profound than fire" extremes, asks Eric Schmidt the awkward questions, and lands on a sober middle. The framing — "we're living inside a trolley problem" — is exactly the spirit of the article: neither dismiss the technology nor catastrophise it, just understand what it actually is.

OpenAI. Four minutes of someone holding up a handwritten linear equation to the camera and ChatGPT tutoring them through it without giving the answer. It is the clearest, shortest demo of "the model can actually see what I'm showing it" and frames the use cases the article recommends — handwritten notes, simple math, captured documents — better than any walkthrough we found.

OpenAI. Two instances of voice mode talking to each other, one of which has camera access to describe the room. Three minutes long and the most efficient way to internalise what makes voice mode different from old "press the microphone, wait, listen" interfaces — interruption, tone, music, real-time vision, all in one clip.

OpenAI. This is the live keynote where ChatGPT's real-time voice mode was first demoed. Mark Chen does the breathing-exercise demo, Barrett Zoph does the math-tutor demo, and then they switch to real-time Italian–English translation. Twenty-six minutes of "ah, that's what they mean by talking to AI like a friend." The model and capabilities shown have only improved since.

Andrej Karpathy. A follow-up where Karpathy actually sits in front of ChatGPT, Claude, Gemini and Perplexity and uses them like a power user. Watch this after the article and his "Intro" — you'll see the mental model put into practice in real time, including the small choices (which model, which mode, when to paste vs upload) that separate frustrated beginners from people who get useful answers on the first try.

Computerphile. A long, calm sit-down with AI-safety researcher Rob Miles on why a model that's trained to predict plausible text will always, structurally, sometimes invent facts. Recorded in early 2023, but the underlying mechanism it explains hasn't changed and the framing is still the cleanest way to internalise "it's a fluency engine, not a truth engine."

IBM Technology. Martin Keen sorts hallucinations into four named buckets — sentence contradictions, prompt contradictions, factual errors, nonsense — and walks through each on a lightboard. After the article gives you the why, this video gives you a vocabulary for spotting the type of mistake in the wild so you can decide how much to trust a given answer.

Tina Huang. A calm 19-minute map of the major model families — OpenAI's GPT line, Anthropic's Claude, Google's Gemini, plus the open-source players — and which tier inside each family is worth your time. After the article tells you "pick one and stick with it for a month," this video tells you what the dropdown menu inside that one is actually offering. Honestly opinionated without being a hot take.

Ryan Doser. Ryan walks through every feature actually gated behind the $20 tier — usage caps, advanced voice mode, image and Sora limits, custom GPTs, the o1 family — and contrasts it with current free alternatives like Claude, Gemini, and Perplexity. The view count is on the lower side because the topic is narrow, but this is the cleanest current side-by-side that doesn't dissolve into hype, which is why we picked it over the louder "Plus changed my life" videos.

Tina Huang. A patient, sidebar-by-sidebar tour of the current ChatGPT app. Useful when the article gets you logged in and you want someone to point at every icon and tell you what it does — voice mode, custom GPTs, projects, memory, the lot — without skipping straight to "advanced prompting."

Matt Wolfe. Recorded in late 2025, so the screens still look like the screens you'll see today. The first third — plan comparison, account setup, dialling in your settings — is exactly the click-by-click ground the article covers, and the back half (projects, canvas, files, branching) gives you a preview of what's actually behind each menu item before you go exploring.

Stanford Online. Andrew Ng, one of the people who actually built the field, talking plainly about what AI is good at, what it isn't, and where the realistic opportunities sit. A useful counterweight to social-media takes — same calm, no-hype tone as the article, just with more depth on the "where is this actually going" question.

3Blue1Brown. The cleanest eight-minute mental model of an LLM on YouTube. Grant Sanderson walks through "next-word prediction," parameters, training, and reinforcement learning from human feedback without ever drifting into hype. Watch it right after the article and the phrase "AI is just autocomplete on steroids" stops being a slogan and starts being a useful shorthand.