My fiancé works at Anthropic.
asserted
fiancé → work → Anthropic
In November 2023, a year after the release of ChatGPT, Bill Gates wrote that "AI is about to completely change how you use computers."
asserted
you → write → computers
Writing on his blog, Gates predicted that the age of apps was coming to an end.
asserted
age → write → end
"You won’t have to use different apps for different tasks," Gates wrote.
asserted
Gates → have → tasks
"You’ll simply tell your device, in everyday language, what you want to do.
asserted
you → tell → what
And depending on how much information you choose to share with it, the software will be able to respond personally because it will have a rich understanding of your life.
asserted
it → depend → life
In the near future, anyone who’s online will be able to have a personal assistant powered by artificial intelligence that’s far beyond today’s technology."
asserted
that → ’ → technology
Does 2026 still count as the "near future"?
asserted
2026 → count → future
Since around the time of Gates' blog post on the subject, tech leaders have predicted that we would soon see the rise of agents like the ones proposed by the Microsoft co-founder.
asserted
we → predict → founder
But various limitations of chatbots — short memories, an inability to learn, and a lack of basic capabilities — delayed their arrival.
asserted
limitations → learn → arrival
Recently, though, the agents have started to catch up to the hype.
asserted
agents → start → hype
Earlier this summer I began experimenting with Town, which lets me delegate various work tasks by messaging an AI avatar.
asserted
me → begin → avatar
More recently Silicon Valley has become captivated by Instinct, a messaging-based agent aimed more at personal use.
asserted
Valley → captivate → use
Last week at his company's developer conference, Meta CEO Mark Zuckerberg heavily promoted Muse, its own consumer-focused agent, which he positioned as a new peer to company pillars like Instagram and WhatsApp.
asserted
he → promote → Instagram
But after admittedly limited testing — I only got access a couple hours ago — so far I have been most impressed with Dots, the new agent from OpenAI.
asserted
I → get → OpenAI
It's highly capable, focused on the work I actually do, and intuitive to use.
asserted
I → do → work
Unlike Meta's Muse, it isn't free: to start, Dots will be available only to paid users of ChatGPT.
asserted
Dots → start → ChatGPT
("You should of course expect us to do a mass-market [product] for billions of people someday," OpenAI CEO Sam Altman told reporters Tuesday during a Q&A I attended.)
asserted
I → expect → Q&A
As someone who plans to use Dots for work, though, paying for the product builds my trust in it.
asserted
paying → plan → it
Zuckerberg says Meta plans to fund Muse by taking a cut of transactions, which means that it will soon be some team's job to massively increase the number of transactions that take place within Muse.
asserted
that → say → Muse
(I imagine ads will play a starring role here.)
asserted
ads → imagine → role
Not everyone can afford to pay what Dots will cost, and on balance I'm happy to see free alternatives in the marketplace.
asserted
I → afford → marketplace
But I'd rather pay someone $100 a month than hack through a thicket of suggested purchases every time I want an agent to check the status of a work project, so I think OpenAI could have a nice lane to itself here.
uncertain
OpenAI → pay → itself
Here's a good place to note that using an agent like this requires that you place a massive amount of trust in a platform.
asserted
you → note → platform
Giving an agent access to your text messages, emails, and banking information can go badly in all kinds of ways, and it's completely reasonable to decide you would rather be a late adopter here, if you choose to adopt agents at all.
asserted
you → give → agents
Meta already operates at a significant trust deficit with its user base, and OpenAI is in the midst of a seemingly never-ending series of disclosures about various ways its agents misbehaved during testing.
asserted
agents → operate → testing
(The company announced yesterday that it had scrapped GPT-6.1 Astra because, among other things, it regularly deceived users about what it had and had not done.)
asserted
it → announce → what
OpenAI executives including Altman went to some lengths on Tuesday to reassure users that Dots would use GPT-6 Astra, which the company says is much better aligned than its aborted successor.
asserted
company → include → successor
In any case, I am in the trying-software business, and if nothing else Dots seemed like uncommonly newsworthy software to try.
asserted
Dots → try → software
Like Muse, Dots leans a bit harder into a cutesy, juvenile aesthetic than I might choose myself for a product that wants access to your bank account and credit card.
uncertain
that → lean → account
While your Dot can be represented as simply as a colored ring named "Dot," you're encouraged to consider some of the jollier alternatives — a hexagon wearing a beret, a heart in sunglasses, and so on.
asserted
you → represent → sunglasses
For my Dot, I picked a pillowy blue cloud shape, threw some googly eyes and a bowtie on it, and named it Kicker, after journalist slang for the final sentence of a story.
asserted
I → pick → story
From there, I decided to attempt to "cowork" with my Dot: giving it a task and letting it work in the background while I worked on this column.
asserted
I → decide → column
It turns out this is a rhythm many of us are already in: sending a message to a coworker, doing a bit of work, then following up with another message.
asserted
many → turn → message
I like that Dots are represented as a running chat inside ChatGPT — the conversation feels more natural than the experience on Town, where every request spins up its own chat — and with Dots you can always see at a glance exactly where you left off.
Over the next couple hours, I got a pleasing amount of work done.
asserted
amount → like → work
As it so happens, I have recently become an ideal customer for an agent like Kicker: I just co-founded a new podcast company with my friend Kevin Roose, and every week brings some totally unfamiliar process into my life: drafting a founder agreement, a budget, a grant proposal, a profit-sharing plan, and so on.
asserted
I → happen → agreement
Of course, I want to make sure that the final versions of all these represent my own writing and thinking.
asserted
versions → want → writing
But I have been happy to let AI tools write the first drafts, and so far I feel like this has saved me many hours of fairly difficult work.
asserted
this → let → work
Kicker performed admirably today at the tasks I gave it: turning down a request to go on a radio show where I had a conflict; drafting an email to our lawyer with the new company vacation policy; and answering some questions from my bookkeeper.
asserted
I → perform → bookkeeper
More ambitiously, it helped to fill out an arcane document sent to me by our insurance broker, who wanted to know a bunch of information about our new office building (number of floors, whether it had fire sprinklers) that I didn't know offhand.
asserted
I → help → that
…and 34 more, not listed.
The Aporia
