From the famous AI 2027 report:
The authors state that predictions beyond 2026 are more uncertain, so I think keeping track on how the earlier predictions are going may tell us more about how the whole scenario works out.
Mainstream narrative shift refers to: "The mainstream narrative around AI has changed from “maybe the hype will blow over” to “guess this is the next big thing” "
Feel free to suggest questions. I will avoid questions that are impossible to resolve accurately (e.g does OpenAI have a secret Agent-1 model aiding their AI R&D)
OpenBrain refers to the placeholder name for the leading AI company in the US in that point in time. CCP refers to the Chinese Communist Party.
I am not sure how easy it will be to assess CAPEX and Power requirements for private companies and AI as a whole so I left these predictions out. But if I am convinced they can be resolved accurately I'll add them
Update 2025-07-07 (PST) (AI summary of creator comment): If the leading AI company is a division of a larger corporation, the creator will use internal revenue reports to identify the leader. The assessment will be based on revenue attributed to AI services (e.g., Gemini or Google AI Pro subscriptions).
Update 2025-08-25 (PST) (AI summary of creator comment): - Late 2026 means end-of-year 2026 (EOY 2026).
"AIs can do everything taught by a CS degree" will be evaluated as: there is an established consensus that AI can fully replace a junior software engineer.
Update 2025-10-22 (PST) (AI summary of creator comment): For the question about OpenBrain's 2026 annual revenue exceeding $45 billion: the revenue will be measured as total revenue from January 1 to December 31, 2026 (not annualized revenue at end of 2026).
1,000Pessoas também estão operando
What are some examples of things typically taught in a CS degree that fable or sol currently can't do? I can not think of any.
@consnop The strongest answer I can come up with is non-core CS subjects that are required or optional in some CS programs, like embedded/FPGA/robotics, public speaking, science classes with labs, etc. The only one of these that's pretty universal across different CS programs is lab science, but even that feels outside the spirit of the question to me given that the actual content of a freshman physics lab is trivial for any frontier LLM.
@MaxLennartson Why? Neither OpenAI nor Anthropic has earned anywhere near $45B in annual revenue yet this year.
Given ARR trends, it’s (very) likely for that to happen by the end of the year, but it hasn’t actually happened yet, and it’s possible something crazy occurs that prevents it from happening.
@MaxLennartson yes, that’s not the same thing as total annual revenue. Given they started the year at ~$9B ARR, they are still well short and well, just re-read my earlier comment
@MaxLennartson remembering that:
the revenue will be measured as total revenue from January 1 to December 31, 2026 (not annualized revenue at end of 2026).
@BayesianTom Essentially every mainstream media narrative, as well as that of the government, has been insisting for over a year that it isn't hype.
Here is there most recent update: https://open.substack.com/pub/aifutures1/p/ai-futures-model-dec-2025-update?r=6lp84s&utm_medium=ios. Daniel and Eli think that the scenario as laid out will basically be the same except that it will happen a few years later. Daniel pointed out that things appear to be 1 year behind the original scenario. As of right now Daniel thinks things will start to unfold in 2029 while Eli thinks things will start to unfold in 2030
Hello @traders with 2025 wrapping up I'll be starting researching into if we have a 1e27 FLOP public model already or not. Appreciate any evidence you can find for each side
@DavidHiggs Hi, sorry for the holdup. I re-read the original article and it says that 45bi is the "projected annual revenue for 2026". I take this to mean total revenue from Jan 1 to Dec 31 (ran it by Claude (pro), ChatGPT and Gemini (free) and they unanimously agreed too)
How will you evaluate "AIs can do everything taught by a CS degree"? What prevents this claim from being true right now?
Also, does 'late 2026' mean by the end of the year, or something else?
@2b3o4o Late 2026 means EOY 2026
I will evaluate "AIs can do everything taught by a CS degree" as "it is established consensus that AI can fully replace a junior software engineer". There are some experiments in that direction right now, but its highly contested/inconclusive.
@JoshSnider I will then try to close the market based on internal revenue reports. E.g if Google wins the AI race then this market closes based on the revenue attributed to AI services, such as Gemini or Google AI Pro subscriptions.