The sight of Russell Crowe’s Oscar-winning performance immediately changes the topic. Me and my interlocutor start swapping random quotes from Gladiator (“AM I NOT MERCIFUL!?!”) and talk about how Ridley Scott definitely should not have made a sequel with sharks swimming in the Colosseum.
But you know who is qualified to answer the question? Google DeepMind’s Demis Hassabis.
Last year, Lex Fridman asked Demis a question about AGI and how the discovery of such a breakthrough could parallel the famous “Move 37”.
A quick refresher: a decade ago, DeepMind created an AI to play the game Go.
It was called AlphaGo. In 2016, AlphaGo took on South Korean Go legend Lee Sedol. During the second game of their match, AlphaGo pulled off a genius non-intuitive play on the 37th move. This move demonstrated AlphaGo’s machine creativity and Sedol was so shook by it that he stepped away from the table.
Lex Fridman: You’ve estimated that we’ll have AGI by 2030, so there’s interesting questions around that. How will we actually know that we got there and what may be the “Move 37” of AGI.
Demis Hassabis: My estimate is sort of a 50% chance in the next five years. So, by 2030 let’s say. I think there’s a good chance that that could happen.
Part of it is what is your definition of AGI? Of course, people arguing about that now and mine is quite a high bar and always has been.
Can we match the cognitive functions that the brain has? We know our brains are pretty much general Turing machines. We created incredible modern civilization with our minds [and that] speaks to how general the brain is.
For us to know we have a true AGI, we would have to make sure that it has all those capabilities. It isn’t kind of a jagged intelligence where some things it is really good at but other things it’s really flawed at. That’s what we currently have with today’s systems. They’re not consistent. You’d want that consistency of intelligence across the board.
Then, we have some missing capabilities. Like the true invention capabilities and creativity that we were talking about earlier. So you’d want to see those. How would you test that? I think you just test it. One way to do it would be kind of a brute force test of tens of thousands of cognitive tasks that we know that humans can do.
Maybe also make the system available to a few hundred of the world’s top experts. The [math genius] Terence Tao of each subject area and give them a month or two and see if they can find an obvious flaw in the system. If they can’t, then I think you can be pretty confident we have a fully general system. […]
I think there are [also] the sort of lighthouse moments like the “Move 37” that I would be looking for. One would be inventing a new conjecture or a new hypothesis about physics like Einstein did. Maybe you could even run the back test of that very rigorously. Have a cut-off [date] of 1900 and then give the system everything that was written up to 1900 and then see if it could come up with special relativity and general relativity, right? Like Einstein did. That would be an interesting test.
Another one would be: can it invent a game like Go? Not just come up with a “Move 37” — [which was] a new strategy — but can it invent a game that is as deep and aesthetically beautiful and as elegant as Go?
Those are the sorts of things I would be looking out for.
This framing of a “Move 37” lighthouse moment is very useful.
Just scrolling through X every day, I frequently see people shocked by some of the current AI’s jagged intelligence capabilities. I’m sure many have felt some form of “Move 37”.
Sure, we are in a hyperbolic hype cycle. There’s also the issue of “I spent 10 hours with a sycophantic chatbot that convinced me I discovered new physics that would make Einstein shit his pants”.
With those caveats, let me tell you the “Move 37” lighthouse moment I’m looking for to see if AGI is here: the major AI labs talk about how they ditched Slack and built an internal messaging tool.
Why? Because communications and co-ordination are massive human bottlenecks…yet the top AI labs (and their thousands of employees) are all running on Slack, which has obvious drawbacks.
If AGI was truly here, it probably isn’t using Slack as the best way to run a cutting-edge AI research organization.
Everyone knows what I’m talking about…the 27-step Slack derangement journey:
-
Wow, this is so much better than email.
- Real-time communication!
- Wow, you can send GIFs.
- I can just DM anyone and make new custom group chats. Amazing.
- Hahaha, there’s a water cooler channel just for #memes. So funny.
- #announcements is an awesome way to get to know your colleagues.
- Can’t believe the CEO just hopped in and started chatting in that Huddle.
-
Dope, you can create channels with external parties to work on projects.
- Slack Connect is next level. The Salesforce integration just pings us sales pipeline updates and my team’s Google Calendar is all sync’d up.
- Whoa, slow down. Stop adding me to new threads.
- Search is awful.
-
Everyone, please put actual announcements in the new #RealAnnouncements channel.
- Mute, Mute, Mute, Mute.
- Think my manager is linking my un-reads to the next promotion cycle.
-
Uh oh, I went on vacation for a week and am now 8,765 messages behind.
- Where the hell is the file I sent?!?!
- These notifications are getting a bit absurd.
- Incredible, another Slack channel created so people can do pretend work.
- We are shutting down #RealAnnouncements. Please direct actual announcements to #RealAnnouncementsOnlyNoReplies.
-
Why I am getting a notification at 10pm on a Saturday?!?!?! My status is DND!!
- Holy crap, they just pivoted the entire company’s strategy under a GIF in a thread from two weeks ago in the #memes channel and I missed it because it was on mute.
- STFU Slackbot.
-
SLACK means “Searchable Log of All Communication and Knowledge”…then why dafuq can I not SEARCH LOGS OF ALL COMMUNICATION AND KNOWLEDGE??!?!?!!
- I quit.
- Man, I hope this new company doesn’t do Slack.
-
[Click Here to Create a Microsoft Teams Account]
- F********CK YOU!!!!!!!
Sorry, where was I? Right. The top AI labs run on Slack even though there clearly has to be a better solution for communication, co-ordination and messaging.
But here we are.
In a piece from July 2025 titled “Reflections on OpenAI”, Calvin French-Owen — who spent a year at OpenAI helping to launch Codex — gave an inside look at how the company operates (side note: French-Owen previously sold Segment to Twilio for $3.2B, which just shows how insane the talent level is at the leading AI labs).
Here was a viral nugget:
An unusual part of OpenAI is that everything, and I mean everything, runs on Slack. There is no email. I maybe received ~10 emails in my entire time there. If you aren’t organized, you will find this incredibly distracting. If you curate your channels and notifications, you can make it pretty workable.
What about Anthropic?
A recent New Yorker profile of Anthropic had 7x mentions of Slack, including how they operated an internal AI-powered vending machine.
While that may not be proof that AI lab runs on Slack, how about this Anthropic blog post from October that announced a new Salesforce integration…and it literally says “Salesforce deploys Claude Code, Anthropic runs on Slack”.
There’s also a bit from the New Yorker piece:
Anthropic’s headquarters, in downtown San Francisco, sits in the shadow of the Salesforce tower. There is no exterior signage. The lobby radiates the personality, warmth, and candor of a Swiss bank. A couple of years ago, the company outgrew its old space and took over a turnkey lease from the messaging company Slack.
It spruced up the place through the comprehensive removal of anything interesting to look at. Even this blankness is doled out grudgingly: all but two of the ten floors that the company occupies are off limits to outsiders. Access to the dark heart of the models is limited even further. Any unwitting move across the wrong transom, I quickly discovered, is instantly neutralized by sentinels in black.
Anthropic runs on Slack and in its old building. So poetic.
Also, the company’s CEO Dario Amodei recently went on the Dwarkesh podcast and talked about how he uses Slack to help keep every employee on the same page vis-à-vis the company’s culture and mission (he spends 30-40% of his CEO time on these areas):
In front of the company every two weeks, I have a three or four-page document and I just talk through three or four different topics about what’s going on internally.
The models we’re producing. The products. The outside industry. The world as a whole as it relates to AI…and geopolitically, in general. Just some mix of that.
I go through very honestly and I say, “This is what I’m thinking, and this is what Anthropic leadership is thinking.”
Then, I answer questions.
That direct connection has a lot of value that is hard to achieve when you’re passing things down the chain six levels deep.
A large fraction of the company comes to attend, either in person or virtually. It really means that you can communicate a lot.
The other thing I do is I have a channel in Slack where I just write a bunch of things and comment a lot. Often that’s in response to things I’m seeing at the company or questions people ask. We do internal surveys and there are things people are concerned about, and so I’ll write them up.
I’m just very honest about these things. I just say them very directly. The point is to get a reputation of telling the company the truth about what’s happening. To call things what they are. To acknowledge problems. To avoid the sort of corpo speak. The kind of defensive communication that often is necessary in public because the world is very large and full of people who are interpreting things in bad faith.
But if you have a company of people who you trust — and we try to hire people that we trust — then you can really just be entirely unfiltered.
For xAI, this article on how the lab employs 900 “AI tutors” says the process is mediated through Slack.
Meanwhile, former Google DeepMind engineer Aleksa Gordić was not a fan of the lab’s use of Slack:
[In 2023, DeepMind] was already bloated (~1200 people, now it’s more like 5000–7000), the culture significantly broken (e.g. memes were being removed from internal Slack channel because they “offended” someone), and I couldn’t freely move around to work on stuff that actually interested me.
As I was explaining earlier, these #memes Slack channels are absolute menaces to corporate productivity.
Everyone knows the pros and cons of Slack. Everyone also knows that if you were to build a corporate messaging and communications platform from scratch in the age of AI, it is going to look different than Slack.
Back in November, economist Tyler Cowen just straight up asked Sam Altman why they don’t build an internal Slack alternative:
COWEN: Someone put an essay online, and it said, in all the time they worked at OpenAI, they hardly ever sent or received an email, that so many things were done over Slack. Why is that? What’s your model of why email is bad and Slack is good for OpenAI?
ALTMAN: I’ll agree, email is bad. I don’t know if Slack is good. I suspect it’s not. I think email is very bad. The threshold to make something better than email is not high, and I think Slack is better than email. We have a lot of things going on at the same time, as you observed, and we have to do things extremely quickly. It’s definitely a very fast-moving organization. There are positives about Slack, but there’s also…I dread the first hour of the morning…the last hour before I go to bed, where I’m just dealing with this explosion of Slack.
I think it does create a lot of fake work.
I suspect there is something new to build that is going to replace a lot of the current office, productivity suite. Whatever you think of Docs, Slides, email, Slack.
[It] will be the AI-driven version of all of these things, not where you tack on the horrible — you accidentally click the wrong place and it tries to write a whole document for you or summarize some thread or whatever — but the actual version of you trusting your AI agent and my AI agent to work most stuff out and escalate to us when necessary.
I think there is probably finally a good solution for someone to make within reach.
COWEN: How far are you from having that internally? Maybe not a product for the whole world — not one that is in every way tested — but that you would use it within OpenAI?
ALTMAN: Very far. But I suspect just because we haven’t made any effort to try, not because the models are that far away.
COWEN: Since talent, time, human capital is so valuable within your company, why shouldn’t that be a priority?
ALTMAN: Probably we should do it, but people get stuck in their own ways of doing things, and a lot of stuff is going very well right now. There’s a lot of activation energy for a big new thing.
So, Sam knows that Slack isn’t the ideal answer. He also knows that OpenAI’s LLM could create a better solution. He thinks they “probably should do it”. But inertia is such a powerful force that doing so would be a distraction.
That was kind of the main point of my article last week (“How Does Docusign have 7,000 employees?”) about why traditional SaaS won’t just disappear overnight in the upcoming AI agent wave.
Employees and their workflows don’t like change.
But Professor Cowen’s point is incredibly valid: “Since talent, time, human capital is so valuable within your company, why shouldn’t that be a priority?”
That’s why I believe a “Move 37” for the arrival of AGI will be some reveal that the leading AI labs are using a new internal AI-powered communication system.
Let’s call it the “Slack AGI Test” (mostly because that’s the title I gave this article before writing it and I figured I should have the title somewhere in the text).
Anyway, OpenAI and Anthropic have already automated a ton of the coding work and are headed to a world in which AI agents create 100% of the code.
AI policy advisor and researcher Dean Ball recently wrote that “the pace of [AI coding] automation will grow during the course of 2026, and within a year or two the effective ‘workforces’ of each frontier lab will grow from the single-digit thousands to tens of thousands, and then hundreds of thousands” because of AI agents.
If you got 6-figure AI agents running around the code base, there ain’t no way that Slack is the best way to manage them.
What is an AI version of a corporate messaging platform — to paraphrase Demis on Go — that is “deep and aesthetically beautiful and elegant”?
There just has to be a better way to co-ordinate. To align intent. To share context. To reduce duplicated work. To prevent mistakes. To make sure no one ever ever ever uses the #announcements channel for a non-announcement. To accelerate decisions. To delegate jobs. To catalogue institutional knowledge. To track progress and keep everyone engaged. To make sure employees stop microwaving leftover salmon in the pantry.
The switching costs of moving away from Slack will be sizeable (including ongoing security, compliance and data retention issues).
But the upside is so clear and AGI should be able to AGI away those challenges.
Having said that, maybe AGI realizes that Slack’s shortcomings are unsolvable and it’s better to leave us humans twiddling around sending memes to each other while it gets the real work done.
I assign a non-zero probability that this is already happening.
In which case…
