In partnership with
It's July 16th, and I sat down to review how many new AI model names had landed since the first of the month.
Claude added Sonnet 5, then Fable 5. OpenAI pushed GPT-5.6 out to everybody. Grok went to 4.5. Then a company I'd genuinely never heard of dropped something called Kimi K3, and I stopped counting.
Multiple names. Three weeks. How can anyone keep up?
If you felt a little behind this month, you weren't imagining it.
Here's what nobody tells you: the names are the least useful part of the whole thing. Every one of those labs used to sell you one flagship — one button, press it, go. Now each of them hands you a family: a big one, a fast one, a cheap one, a specialist. You open a tool you've used for months and there's suddenly a dropdown where the button used to be.
So you freeze. Not because you're slow. Because someone replaced a button with a menu and didn't explain the menu.
THE QUICK TAKEAWAY 💡
You are not choosing a model. You're choosing a job.
The dropdown exists for people who want to tune things. You just need to know what you sat down to do — and the tool will almost always pick a sensible default for that on its own.
Names change monthly. The job you came to do doesn't.
Start with the job, not the name
Four jobs cover almost everything a beginner actually needs. Pick the job first, and the tool question mostly answers itself.
You need to write something, or think out loud. Emails you're avoiding, a report, a hard message to a coworker, a half-formed idea you want to argue with. This is where I use Claude, and it's not just my preference — the 2026 comparison roundups keep landing on the same finding, that it handles natural prose, tone, and long documents better than the alternatives. When I drafted the layoff survival guide, I didn't ask it for "a guide." I told it who I was writing for and what I wished someone had told me, and worked from there.
You need a quick answer while you're doing something else. Looking something up, checking a fact, "what's another word for." Honestly? Use whatever is already open. This is the job where the tool matters least, and where people waste the most time deciding.
You'd rather talk than type. This got real this month — ChatGPT's voice mode became the default for everyone, free tier included, with live translation built in. If typing is the thing that's been stopping you, that's the on-ramp. No prompt writing, no menu. You just talk.
You need a multi-step errand done, not advice about it. That's an agent — which is exactly what I wrote about last week.
The AI Work Handbook That Cuts Your Workday in Half
The 8-hour workday is becoming a 4-hour workday for people who know how to use AI.
Everyone else is still catching up.
This AI work playbook shows you exactly how to cut your work hours in half using AI.
Sign up for Superhuman AI and get:
50+ step-by-step AI tutorials to cut your workload in half — covering every part of your workday, from emails to strategy, used by 1M+ professionals at Google, Microsoft, and NASA
Superhuman AI newsletter (4 min daily) so you keep discovering new AI tools and skills to stay ahead in your career — the playbook is just the start
Which brings me back to last week
Last week I said the word "chatbot" had stopped fitting, because I'd watched Claude take a messy pile of notes and hand me back a finished weekly plan instead of instructions for making one.
Here's how that connects: an agent is just job #4 on that list. It's not a new species of AI you have to go study. It's the same tools, doing several steps in a row instead of one.
Which is the whole point of the job-first approach. When the next name lands — and one will, probably before this email gets stale — you don't have to research it. You ask one question: does this help me with a job I actually have? Usually the answer is no, and you get to keep your afternoon.
And no, you probably shouldn't pay for it yet
One more thing, because it keeps coming up.
Every one of the four major tools — Claude, ChatGPT, Gemini, Grok — has a free tier that is genuinely good now. Not a crippled demo. Good enough that most people reading this could do everything in this email without entering a card.
Learn on the free tier. Pay when you hit a wall you can name. If you can't name the wall, you haven't hit it.
BEHIND THE SCENES 👀
Here's your homework, and it's small on purpose.
Pick one job from that list of four — just one, the one you'd actually use this week. Do it in Claude's free tier five times over the next seven days. Same job, five times.
That's it. Don't switch tools. Don't read a comparison article. Don't touch the dropdown.
Five reps on one job will teach you more than five months of keeping up with names. Because the skill you're building isn't "knowing the models." It's knowing how to ask well — and that skill survives every rename, every version bump, every new lab that shows up out of nowhere on a Monday.
You're earlier than you think, not later.
Spoiler: I meant what I said last week — what I write next comes from your replies, and a few of the questions I'm sitting on could each fill an issue on their own.
So tell me: what should I cover next? Hit reply with the AI thing that's been quietly confusing you — the term you nod along to, the button you've never clicked, the thing you'd be a little embarrassed to ask out loud. Those are my favorite ones to write. No question is too basic here; the basic ones are the ones nobody bothers to answer properly.
Anna
