Companies are not investing enough AI build time in process management. They're too focused on product or individual productivity, and they're missing handoffs. When I finish my meetings, all of my followup messages are already written in Slack for me. I tell someone I'll pull them the recent client numbers for something, and after the meeting, the note is already drafted with the numbers in Slack. I review and send. Or I dictate a social post during a content call, and after the meeting, the post is already saved in Notion and, fully tagged and ready. Think about the handoffs!

The easiest AI starting point I know, for ANYONE staring at a blank chat window: Click the microphone. Complain for a few minutes about your business, your calendar, your kids' school schedule, all of it. Then say "help, interview me." Voice is about 4x faster than typing, and you can do it on a walk. I do a lot of my AI work on 45-minute walks around New York. My team is currently working toward "no-screen" days - a fun challenge to take on if you're looking for one. This workflow alone gets you to about level three of the five proactive AI levels I teach.

Presumably, many people who bought Mac minis would have been willing to spend that same amount (nearly $1k) for Codex/Claude Code to be always on and in the cloud. Possibly even more if there is zero-headache setup.

Posting online used to be more fun. Now it’s 50% AI-generated replies asking me a nonsense follow-up question and thousands of AI agents scraping my tweets to help others research or write. I find myself sending more curated email blasts because the replies are incredibly human and rich, as well as holding or attending more live meetings/events and in-person engagements. Considering also just sending a 2min voice memo out to all my followers most days. Not a polished formal podcast. Just a “inside the corners of Allie’s mind” that doesn’t require me to get on camera.

"As [Fortune 500 companies] are testing new features, they are taking those learnings and bringing them to everyone else. That breakthrough becomes everyone else's starting point, and Slack is where all of these companies are making that connection." - @alliekmiller #DF26#

Whenever I see a new voice-first interface, a tiny voice inside of me screams with joy. Voice is the interface of the present and future. I see you, Google. Do your thing.

To the people quitting their jobs and making proclamations of AI concerns, here is what you should ACTUALLY be sharing in your tweet. ⬇️ - how long you have worked in AI - of that time, how long you have been worried (and specifically about what? Misalignment? Bioweapons? International coordination?) - your magnitude of worry and why (including timelines, confidence scores, assumptions you’re tracking) - actual specific things that can go wrong (“kill us all” is the most vague panic-inducing option you have; your previous employer would not have hired you if that was your answer on AI risk examples - pretend you are getting interviewed because Anderson Cooper likely will do it anyways), you can hedge, you can say “and that’s just what we can imagine today” or “and that’s just with the next model family” but say a real thing that non-AI-pilled-superusers can actually picture - you need specific things you are calling for. Real action. A literal meeting is not enough. Your AI risk mitigation suggestion against “AI killing us all” needs to be more impressive than a parent-teacher meeting. Again, you are smarter than your current tweets and you are wasting your opportunity to drive change just like Hinton when he left Google. Examples include: environmental standards at data centers (air pollution, water quality), third party evaluators in labs, increased federal investment in open source, verifiable criteria to join early testing groups, whistleblower protections, required public risk disclosures before model releases, 3x minimum increase in federal renewables investment to catch up to China’s per capita spend, block chip exports to non-allies, highlight companies doing great AI safety work that more people should support, literally anything other than just “guys idk we gotta meet or something” - have a non-engineer review it - have a person not in tech review it - have someone outside of your generation review it - have at least one AI optimist review it - and this one may sound odd but I think you should read one war-time presidential speech before it You have an opportunity to do something deeply meaningful on your way out. That is more likely than not one of the reasons you joined your company in the first place. Tweeting vagueties and inciting panic contributes the opposite.

If China “beats” the US in the AI race, I’d bet two big reasons would be their outsized focus on open models and their massive renewable energy buildouts. China added 11x as much renewable power capacity as the US in 2025. Even adjusting for population differences, it’s still almost THREE TIMES our added capacity per person. This potential rollback from the EPA is sending us in the wrong direction.

Perhaps it is worth stating that if the AI labs cannot organize what is essentially a fancier friend hang (a gathering of the top minds, representatives from all the labs + research institutions like METR + federal agencies, multiple days/weeks, a review of the recent cyber incidents, review of threats including cyberwarfare/biological warfare/electric grid interference, the beginnings of an action plan), then we should not trust them to do whatever comes after that. Get the best celebrity wedding planner in DC, pick a date, and figure this out. I can’t read another tweet from a well-meaning researcher saying “we have to get together and talk”. Ok then. Pick a date. 11-11 sounds good. It’s a Wednesday. And it’s the same for both the US and Europe, so there will be no confusion, even better. It emphasizes there is one shot to getting this right.

I showered with my AI, and it was a horrible mistake. As the world gets more AI-friendly, and as we find more and more things to be productive about (and I don't want everything in my life to actually be productive), I have been testing and experimenting with AI usage in different contexts. And Codex live voice mode opened up new options for me. One time, instead of waiting to clean my apartment at the end of the day or on the weekend, I actually did it during the workday while I was dictating to Codex and processing tasks that I had to do. Another time, I had to fill out this form for my Apple Business account. And Codex did it all for me - it even found my W-9 on my laptop to pull my EIN. It completed the whole form so that I could run errands. Instead of being strapped to my computer and having to dig up all these files and numbers on my own, I could do double duty and get personal things off my plate without losing time. I use AI on walks. And importantly, it's not that I'm adding AI to the walks that I'm already doing (for that, I still prefer Sam Cooke or Hello Puzzlers). It's that I'm able to TAKE MORE WALKS that I wasn't able to before because of AI. So now, I might go on a walk at 4pm and process all the Slack messages that I didn't respond to or review a client contract. As a team, sometimes we even take team-wide walk breaks to work 1:1 with our AIs. BUT. I just tried one new workflow that I hated with the fury of a thousand suns. And that was trying voice mode while I was in the shower. It was effective in the sense that the AI could hear me well and could process tasks (unlike the Apple AI assistant who shall remain nameless, who can't hear me if a single pin is dropped in the background). Sure, there was a little annoyance when the phone had a permission pop-up and I had to reach my hand out of the shower to press it. But my biggest gripe was that I just hated the overall experience. Hated it. There are some parts of my life that I really don't want to add AI into right now. And it's really in those times where I allow my brain to dance off into the edges of my imagination, or truly relax and take time away from execution. And I just want this to be a reminder that we need space to be away from tech and just think and exist. That could be on a nature hike, or in the shower, or my favorite meditative place in the whole wide world (the middle of a car wash). There is no perfect optimization formula that is telling us that every minute of every day has to be AI-enabled. So I will no longer be showering with my AI and will stick to my rubber duck. Thank you.