Long Live QWERTY

Long Live QWERTY

Picture the last time you used AI. You were probably at a laptop, or thumbing something into your phone. A screen, a keyboard, a text box. That's been the whole experience for almost three years now.

A lot of very rich companies are betting that's about to change. Not the AI part. The where part. The screen and the text box.

The question I keep chewing on is a simple one, and it matters if you run a business: a few years from now, when you "use AI," what are you actually doing? Talking to a pair of glasses? Steering it with a little control pad? Just chatting into your headphones all day? Something nobody's built yet? Or are you still sitting at a laptop, same as today?

This week:

The real question under all the AI hardware hype, and my honest guess at where it lands

A spotlight for the nonprofits reading this: how to stop burning weeks hunting for grants you could actually win

Anthropic made a call on its best model that surprised me, plus the biggest open-source AI release yet

The Deep Dive

Long Live QWERTY

Last week I wrote about Apple suing OpenAI. Quick recap in case you missed it: Apple took OpenAI to court, accusing it of stealing trade secrets to build a competing hardware business. The two of them are partners and suing each other at the same time. Wild stuff.

But the lawsuit was never the most interesting part. The interesting part is why two of the biggest companies on earth are willing to go to war at all. They're fighting over a thing that doesn't exist yet. And the reason they care so much is the question I actually want to sit with today.

Here's the setup. For almost three years, "using AI" has meant the same thing: you go to a screen, you open a tab or an app, you type. The phone and the laptop have been the front door. Everyone's now racing to build the next front door. Let me lay out the contenders, and then I'll tell you where I'm actually placing my bet.

Door number one: glasses. At CES this January, the whole show pointed here. AI that sees what you see and hears what you hear, no screen required. Meta's Ray-Ban glasses are the first pair normal people wear in public without feeling like a robot. Samsung unveils its Galaxy Glasses this week, and Apple's own pair is reportedly a couple years out. The whole industry is clearly circling this idea.

Door number two: a physical control pad built just for driving AI. OpenAI actually shipped one of these. It's called the Codex Micro, a $230 little slab of knobs and keys, and it's the most interesting AI device I've seen precisely because it's so unglamorous. There's a joystick you flick to fire off a task, dedicated keys to accept or reject what the AI did, and a dial you literally turn to crank the AI's brainpower up or down depending on how hard the problem is. Each key even glows a different color to show you what your AI agents are doing, this one's thinking, that one's waiting on you. It's a cockpit for AI. Not a screen replacement, a steering wheel.

Door number three: headphones, and the whole interface is just talking. No screen, no glasses on your face, nothing to look at. You put in an earbud and you have a conversation, ask a question, get an answer, back and forth, like the AI is a colleague on a call that never ends. A lot of people already half-live here with AirPods; this just makes the AI the thing on the other end.

Door number four: something none of us have thought of yet. This is the one I don't want to skip past, because it's historically where the real winner comes from. Nobody predicted the iPhone by staring at BlackBerrys. The device that actually replaces the laptop might be something that isn't on any show floor right now, that we'll all pretend we saw coming.

And door number five, don't laugh: nothing changes. You keep using a laptop and a phone. The AI just gets quietly better inside the tools you already have.

Here's my honest bet, and I'll be straight with you: I'm putting my chips on door five or door four. Status quo, or the thing we haven't imagined yet. Not the glasses, not the control pad, not the headphones.

Why? Because none of the options actually on the table right now feel enough better to move the general public. Every "iPhone killer" of the last fifteen years promised to replace the phone, and none did, because the phone is genuinely great and the alternatives were only a little different. Remember the Humane AI Pin? It tried to replace the phone outright and became a paperweight when the company folded. Glasses are cool but most people won't wear a computer on their face all day. The control pad is genuinely useful, but it's for coders, not your accountant. Headphones-only is great until you need to actually see something. None of these clear the bar of "so much better that everyone switches." The graveyard is full of gadgets that were sure they'd win.

Now I'll push back on my own bet, because I don't want to be the guy who said the iPhone would flop. There's a real difference between reading AI on a screen and actually operating it. That Codex Micro is a small thing, but it's pointing at something real: as AI shifts from "answer my question" to "go do this whole task for me," you stop typing at it and start steering it. And steering wants controls, a dial, a button, a light that tells you what's happening. If your team ends up running five AI agents at once the way a pilot runs a cockpit, the screen-and-text-box setup starts to feel primitive fast. So I could be wrong. The door I'm calling "status quo" might quietly turn into something new without anyone noticing the switch.

But here's what it means for you, and it's the practical part I actually want you to walk away with. Don't bet your money, or your habits, on the laptop and the smartphone going away anytime soon. I wouldn't. So keep learning to type. Keep getting better at saying exactly what you mean, because whatever door we walk through, you're going to be talking to these things in words, and the person who can say precisely what they want will always get more out of AI than the person who can't. Prompting is just being articulate. That skill doesn't expire when the hardware changes, it's the one thing that carries across every door. Glasses, control pad, headphones, or the same laptop you've got open right now, the winners will be the people who know how to ask well. That's the part we help businesses get right, and it has nothing to do with what gadget you end up holding.

From the Field

For the Nonprofits Reading This: Stop Burning Weeks Chasing Grants You Were Never Going to Win

This one's for the nonprofit folks on this list, and I know there are a few of you.

Here's the part of grant work nobody warns you about. It's not the writing. It's the hunting. You know the money is out there somewhere, so you spend days, sometimes weeks, digging through databases and funder pages trying to figure out which grants are actually worth your time. And most of what you find is a dead end: too big, wrong region, wrong focus area, or a foundation that only ever funds groups three times your size. You can pour a whole week into research and come out the other end with two real leads. That time is the expensive part, and it's time you don't have.

There's a Claude connector called Granted that collapses that week into a conversation. You turn it on, and then you just ask, in plain language, right inside a normal Claude chat. "Find grants a small arts nonprofit in Arizona could realistically win, with deadlines in the next 60 days." "Which foundations near us actually fund youth programs at our budget size?" It searches a live database of tens of thousands of grants and 133,000 foundation profiles and hands back real, scored, ranked results, not a generic web summary. The key word is realistically: it'll tell you whether a funder has a history of backing work like yours and at your scale, so you're not writing proposals into a void.

A human still has to interpret the results and the tool doesn't do the relationship part, and it shouldn't. But going from a week of dead-end research to a short list of grants you could actually land, in one sitting, hands you back the hours that matter most.

Why this travels: every organization, nonprofit or not, burns real time hunting for the right opportunity in a pile of wrong ones. Pointing AI at that search, so your people spend their hours on the ones you can win instead of finding them, is one of the highest-return things you can do with it.

AI News Worth Your Time

Three Stories From This Week

Anthropic kept its most powerful model in the premium plans instead of pulling it

Anthropic had planned to take its top model, Fable 5, out of subscription plans and make you pay by the token to use it. This week it reversed course: as of today, Fable 5 is a permanent, included part of the Max and Team Premium plans, up to half your weekly usage.

My take: this surprised me, in a good way. The honest tradeoff is that Pro and lower-tier folks got pushed toward pay-as-you-go, so it's not a win for everyone. But the top model staying bundled into the premium plans, right as cheaper competitors are circling, tells you Anthropic decided keeping its heaviest users happy was worth more than squeezing them. If you're building your work around a specific model, that's a decent signal it'll stick around longer than the usual tech treadmill trains you to fear. READ THE REPORT →

The best open-source AI model ever just dropped, and it's free

A Chinese lab called Moonshot AI released Kimi K3, the largest open-source model ever built, and on the benchmarks it runs neck and neck with the top paid systems from OpenAI and Anthropic. The full weights go public July 27th, which means any company can, in theory, download it and run it on its own machines.

My take: watch this one closely. There's a running debate about whether these Chinese labs are genuinely innovating or just cleverly copying the American frontier models, and honestly, it doesn't matter much for you. Copying or not, a free model this capable is seriously impressive and puts real pressure on the big names. The gap between free open models and the expensive frontier ones is now the smallest it's ever been. For most small businesses the paid tools are still the easier, safer bet, you're paying for the polish and the support. But cheap, capable, self-hostable AI puts real pressure on prices across the board, and that's good for everyone buying. READ THE VENTUREBEAT REPORT →

The head of Google DeepMind wants a referee for AI, and he wants it by year-end

Demis Hassabis, who runs Google DeepMind and won a Nobel Prize, published a plan this month calling for a U.S.-led body that would safety-test the most powerful AI models before they're released to the public, checking for things like the ability to help build bioweapons or run cyberattacks. He wants it up and running before the end of the year, and he floated the idea that if a model turns out to be too dangerous, the labs could coordinate a slowdown.

My take: this matters even if you never touch a frontier model. After the chaos earlier this summer, when the government scrambled to ban a couple of models overnight, this is the clearest sign yet that the "move fast and figure out the rules later" era is ending. For a business, that's mostly good news, more oversight means fewer nasty surprises baked into the tools you rely on. Worth knowing which way the wind's blowing. READ THE REPORT →

That's it for this week.

If this was useful, forward it to someone who runs a business and keeps hearing "AI hardware" in the news without knowing whether it's hype or something they should actually care about. That's who this is for.

And if you want to talk about what AI could do for your business, not in theory, but in your workflows, with your team, on your timeline, reach out at [email protected]. That's what we're here for.