Allie K. Miller

#1 Most Followed Voice in AI Business (2M followers). Former Amazon, IBM. Time100 AI. Fortune 500 and startup AI advisor, public speaker. AI-First courses in 🔗

[Allie K. Miller @alliekmiller]

There's a difference between using AI and building with it.

Copy-pasting ChatGPT prompts will only get you so far. I want to help you learn to build personal AI software, automations, and tools that actually solve your problems.

The AI Fast Track is a free 5-day course. Tens of thousands of people have taken it.

Sign up and get it in your inbox tomorrow: https://t.co/9PlxHjTj4T

[Allie K. Miller @alliekmiller]

OpenAI launched Dots at DevDay.

My thoughts.

First, the good. Always-on is really really really what the market needed. It's why I find myself reaching for Instinct so often. I don’t have to think about WiFi or my computer being open or Mac mini being on.

Second, every Dot gets its own dedicated VM. Clearly a big compute commit, and it's the same directional bet Meta made with Muse. Will gladly take a persistent dedicated environment over a just-in-time spin-up (persistence is what makes an agent feel more like a teammate).

Which unlocks the overarching use case: AI as an ambient watchdog. "Watch my Slacks and tell me to step in if two teams are duplicating work" or "watch my spend and apply for a better credit card” or “keep an eye on my calendar and handle all gift ordering” or “watch my Apple Watch and manage all my gym class bookings”.

Dots give AI *ongoing responsibilities* 🔍

Underneath it all, the architecture feels like it’s settling on a direction too. A tax exec recently asked me about governance, and between smaller models and the Decisions API, what's winning right now feels like a big brain managing a thousand tiny ones (versus just one big brain).

Also - anyone else notice the whole Dots product video was dictation? And in fact, not a single computer screen was shown? Some phones, tablets, TV monitors, and the back of one laptop. In my mind, this is a device-agnostic pitch. Maybe Dots are the brain or the working paradigm for the Ive device. Also, typing is dead.

Multitasking - hypertasking, really - still feels unsolved, even if that’s a future state for Dots. A proactive agent means YOU'RE the one context-switching now, and I don't think we've cracked that UI. I hate chat threads as the main option. It's why so many dev superusers create their own work interface. Designers, what a great opportunity for you!

The name... I’m sure they argued over names nonstop. "Dots" sells the swarm, but it makes it sound like you have to direct-manage a whole fleet of helpers. Although OpenAI is only giving users one primary Dot atm, the “team of dots” is promised for the future. Product name also sounds 10% too cute for the workplace but it’s also a response to how dev-oriented “Codex” felt.

It’s a bit buggy right now, the early reviewers are right. But expect that to change over the next few weeks - this was an intentional early announce to pull attention back off of Muse, GrokBot, and Instinct. Smart imo.

Always-on, watching, acting before you ask.

The word of the year continues to be: PROACTIVE.

[Allie K. Miller @alliekmiller]

Lisa Su has always been a big player.

The OpenAI chip deal last fall sent AMD stock up ~25% in a day - and just in the last week she:

1) announced an $8.2B deal to acquire Fei-Fei Li's World Labs
2) made AMD the first woman-led company ever to cross $1T
3) got seated 4 away from POTUS aka closer than Broadcom and the cyber firms

[Allie K. Miller @alliekmiller]

@danshipper @sama When did you start using the word “yall”?

[Allie K. Miller @alliekmiller]

By far one of the most interesting parts of Dev Day is - demo failures or not - those folks attempted to voice dictate EVERYTHING.

If you’re not using voice AI, it’s clear you should.

Future products, features, and experiences will be based around it.

Time to get that DJI mic! 🗣️

[Allie K. Miller @alliekmiller]

Another fuzzy blobby sheep!

Or frog.

Count me in.

[OpenAI @OpenAI]

10am PT, on the dot.

[Allie K. Miller @alliekmiller]

OpenAI just announced the $200 Pro plan is back but with half the total API spend included, offset by some of their smaller (NOT Astra!) getting a 50% price drop.

Does this impact the average user? No.

Recent superusers? Yes.

Why? Astra usage.

We are still VERY much in a compute-constrained world, and here are the clues that point to that. When many of these start changing (in a year, in two decades, who knows), you can assume we’re moving out of a compute-constrained world. Until then, tokenomics matter.

CLUES:
- limits on new subscribers at higher tier subscriptions
- usage limits by window or week
- techies talking about the future promise of abundance (aka it’s not here yet)
- model-specific limitations (like fable)
- slower responses from models
- default medium thinking
- default automode
- autorouting to smaller models
- CEOs tweeting “our GPUs are melting”
- free tier users not getting frontier models
- labs signing big energy deals
- efficiency breakthroughs like deepseek moving markets (Nvidia saw a big drop, remember?)
- hundreds of billions of dollars in future compute commits from both Anthropic and OpenAI
- labs literally saying they’re compute-constrained
- not seeing “unlimited” allowances everywhere
- hearing about GPU shortages at cloud providers

Also, dropping the hammer on usage right before a ton of (likely great) press coverage - smart timing - even if the artificially inflated API price idea sounds like a strawman.

[Tibo @thsottiaux]

Hi,

Tomorrow we are re-opening the Pro $200 subscriptions to new subscribers, but together with it we are also changing how we calculate the usage for it. In effect, if you do the math, it will net out at half the dollar in API spend compared to the old Pro $200 plan.

Now that it's said, let me explain why this is happening and why you will still get more work done than if you were on the Pro $200 subscription one month ago.

(a) We didn't want to compromise in other ways and are committing to not reintroducing the 5h limit, so that you can fully use the weekly usage when you want.
(b) On the subscription, we guarantee that over time you always get more work done and with an increasing level of quality. This means that you will continue to get more value per dollar spent as a result of models getting more efficient and us passing down the improvements in the form of API price reductions.
(c) We don't want to put an incentive on ourselves to artificially inflate the API list prices to make it look like you are getting a lot (and workaround it through discounts, etc). Instead we want to continue to both rapidly reduce prices and increase capabilities of models on the API. This week we introduced GPT-6 Sol and GPT-6 Luna at 50% of their previous price. Over time, we see prices go low enough that it makes sense for most to buy usage as needed without there being a significant gap between what you get in a subscription and what you get in the API for a dollar spent.
(d) Tomorrow, we are adding more things to the subscription that won't draw on the usage, I won't reveal what that is yet.

I wanted to be transparent before all the big announcements tomorrow. Lots of new exciting things are coming to the subscriptions that will make it super compelling, but I wanted to make sure to share this change ahead of time so you can all understand it before we shower you with good news.

Codexingly,
Tibo

[Allie K. Miller @alliekmiller]

@HarperSCarroll @petergyang Astra would never 😇 and right back at ya!

[Allie K. Miller @alliekmiller]

@petergyang Photo of your intestines or it didn’t happen

[Allie K. Miller @alliekmiller]

Let's do the Min Choi thing and dig into 19 interesting Meta Muse examples.

The good and the bad.

First, the good.

[Allie K. Miller @alliekmiller]

That's where we are.

Saving thousands of dollars and also sharing our home addresses with strangers.

Remember that all of these stories are anecdotal, your mileage and risk tolerance and love of Zuck may vary, and we might learn more about each of these use cases as time rolls on.

What I'd bet on: every AI lab is gunning for proactive AI agents. Muse and Instinct are first movers. The others will follow.

[Allie K. Miller @alliekmiller]

This user's Muse sent an eerie photo that looked like it came from someone else entirely.

Not clear whether it's literally from another user (and how that would happen) or whether the AI hallucinated the image.

https://t.co/QIXb9TnTi9

[Jules Forrest @julesforrest]

Alex HW, if you exist, Muse is leaking your CompTIA practice exam study session!

[Allie K. Miller @alliekmiller]

Muse prioritized keeping the phone quiet over the user's previously requested medication reminder.

Poor tradeoff imo.

https://t.co/IXOvfp1Ndi

[DannyDuberstein @DrPaulProteus_]

Tough look from @Muse. Set a reminder to take my medication ahead of schedule and to reserve a car for my early morning flight and it was apparently going to silently fail. Building agents is very hard but this kills user trust.

[Allie K. Miller @alliekmiller]

Another user @MattRobbt, trying to sell keyboard keys on FB, reported that his Muse agreed on a lowballed price he never approved, told a stranger on FB Marketplace his home address, and lied and said he was home when he wasn't.

Scariest example I saw.

https://t.co/QRlXXqtvyq

[Ray Wong @raywongy]

Deleted Muse after seeing this post on Threads about how it told some Facebook Marketplace sellers the guy’s address and they showed up at his door

Dangerous and creepy

This would have been 1000x worse if the person was a woman

[Allie K. Miller @alliekmiller]

Users have hilarious screenshots of call transcripts.

Another user @PaulJLipsky said, "Half the companies hang up on my agent."

Great if you're casting a wide net or trying for a price reduction on something relatively inconsequential, but wouldn't use on high risk asks.

https://t.co/ada3OCx1Sa

[Ankit Gordhandas @AnkitGordhandas]

@petergyang @PaulJLipsky @Muse Yeppp!