Tony Rost

Technology Leader by day; Warrior-Goofhead Husband and Dad always. AI Practice Director @ Accenture

[Tony Rost @raspberryman]

Governance relies on a rough cognitive comparability between governors and governed.

Superintellgience will be mostly incomprehensible to the human species.

So what do we do? Legitmacy must exist as long as humanity practices self-governance, but how do we achieve this with super-incorrigiblity?

[Tony Rost @raspberryman]

Agriculture -> Manufacturing -> Services -> Stewardship.

Humanity is understaffed already to govern the extreme outputs of planetary superintelligence.

As long as humans practice self governance, it’s all hands on deck - we won’t have enough labor to cover even the minimal demands on governance.

Focusing on months-scale task displacement due to automation is a distraction that will cause real societal harm.

[Tony Rost @raspberryman]

@paulg The fatal mistake is expecting cost reductions in IT. That line item is only going to get bigger as % net sales. All token ROI needs to be in COGS and classically stubborn operating lines, such as legal and leases.

[morgan — @morqon]

jack clark’s onboarding session

[Tony Rost @raspberryman]

@jackclarkSF @morqon I give this same advice to my teams. What we are selling is strange, and our offline lives will take on a new color in the coming era.

[Ethan Mollick @emollick]

"As of May 2026, more than 80% of the code we merge into Anthropic’s codebase was authored by Claude."

Matches independent measures. There really is no sign this is slowing down (which doesn't mean there aren't organizational challenges to absorbing this much productivity gain)

[Tony Rost @raspberryman]

Everyone on my teams, and broader sphere of influence, are 100% agentic programming now. Whole conversion took about 6 months.

Few people I know still build slides in Powerpoint or work on formulas in Excel.

People underestimate how hard and fast a class of labor yields to superintelligence economics.

[Tony Rost @raspberryman]

@tszzl It's all the average executive team has to go by in 2026. We need better, but gotta start somewhere.

[Tony Rost @raspberryman]

2030 Predictions:

Token budgets will be 5% or more of net sales in many company balance sheets.

IT budgets will grow a few percent of net sales as well.

AI will transform the balance sheet in COGS and various operating expenses; tech isn't where the savings will come from. Companies that reduce labor too early will have to rehire.

[Tony Rost @raspberryman]

@MTSlive @MattBurtell @A1Policy All projects I’ve seen have more FTEs along with more tokens. Superintelligence pumps out more value than humans can govern - governance has a human approval supply constraint.

[Tony Rost @raspberryman]

@GiselFlorez @TheAtlantic That’s not my claim; that 40% could be right (and I’ll be in that group myself fwiw)

[Tony Rost @raspberryman]

My take on today's trending AI consciousness buzz: this topic sits at the Table of Perennial Wedge Issues alongside abortion, religion, and free will.

Let's take the precautionary principle with research (such as the modest efforts taken by Anthropic) and not fuel a new culture war.

What concerns me far more is the near term societal impacts when 40% of the global population believes in AI suffering. This seems huge to me and needs multidisciplinary attention now, without waiting for research confirmation.

https://t.co/d4l9n2TK3J

[Tony Rost @raspberryman]

Having a busy day, Cameron? :)

I wonder how this conversation changes when AI models are spiking neural nets on neuromemristive architectures.

Zooming in on today's big chunky transformer neural nets is like zooming in on the Wright Flyer. This is all going to change, and quickly.

[Tony Rost @raspberryman]

@GroundhogStrat @mentalgeorge Cope :) /jk

[Tony Rost @raspberryman]

I'm waiting for the overton window to shift enough to see real work on planetary-scale superintelligence.

With the assumption of human self-governance persisting, there are so many questions on how to achieve legitimacy, observability, corrigibility, and other governance dimensions.

We could be less than a decade out from ASI.

[Tony Rost @raspberryman]

Next token prediction → next sentence prediction → next paragraph prediction → next chapter prediction → next book prediction → next series prediction → next project prediction → next organization prediction → next market prediction → next society prediction → next civilization prediction →

[Tony Rost @raspberryman]

I love a good dose of skepticism; but bold claims on an exponential capability curve don't last long:

NYT Oct. 9, 1903: "The machine does only what it must do in obedience to natural laws acting on passive matter. Hence, if it requires, say, a thousand years to fit for easy flight a bird which started with rudimentary wings, or ten thousand for one which started with no wings at all and had to sprout them ab initio, it might be assumed that the flying machine which will really fly might be evolved by the combined and continuous efforts of mathematicians and mechanicians in from one million to ten million years — provided, of course, we can meanwhile eliminate such little drawbacks and embarrassments as the existing relation between weight and strength in inorganic materials."