Sozzi Unleashed

Mike Schroepfer on choosing AI specifics over absolutes

Mike Schroepfer· Founding Partner at Gigascale Capital
·~8 min·English·Yahoo Finance
AgentsAI SafetyCybersecurityBusiness Strategy
TL;DR

Former Meta CTO Mike Schroepfer makes a pragmatist's case: the always-on Muse assistant is already useful and headed for the enterprise, and the AI-risk conversation should trade sci-fi absolutes for the concrete near-term problem -- cybersecurity, where an autonomous attacker only has to find one unlocked door.

01The Shift

The Assistant That Never Logs Off

Schroepfer's case for Muse is that it is always on and proactive -- a standing assistant you delegate to, not an app you open and close.

it is always on 24/7, proactive. It is just a better way to interact with these AI systems.

— Mike Schroepfer, Sozzi Unleashed
Key Insight
The jump he is describing is not a better chatbot but a change in default: AI moves from a thing you consciously open to a process that is already running and surfaces work on its own. Once that is the baseline, 'use AI' stops being an action and becomes an ambient layer -- which is exactly why he expects enterprises to want it next.
02The Unlock

The More It Connects, the More It Does

Schroepfer says Muse is already useful just driving a web browser -- and that once it has direct API connections into Slack, your calendar, payments and the ERP, it becomes far more useful still.

once it has a good API connection into everything, it's going to be so much more useful.

— Mike Schroepfer, Sozzi Unleashed
Key Insight
He is pointing at where enterprise value accrues: the breadth of systems the assistant can reach. Muse is already useful just driving a browser; direct API connections into your tools let it act in far more of them. Usefulness grows with reach.
03The Frame

Specifics Beat Absolutes

Schroepfer argues the AI-risk conversation gets easier to engage with when you drop the grand absolutes and name a concrete risk together with its response.

When we talk in sort of big absolutes and generics, it's hard for anyone to engage.

— Mike Schroepfer, Sozzi Unleashed
Key Insight
Specific framing turns a broad worry into concrete questions -- what is the risk, what is the response, is it worth it -- which is what lets people actually engage and weigh the tradeoffs. Naming a risk precisely is the first move toward dealing with it.
04The Real Risk

The Attacker Only Needs One Door

The near-term risk he actually worries about is cybersecurity, where the math favors the attacker: the defender must lock every door, while an automated agent can patiently try them all.

it's way harder to defend than it is to attack. I got to lock every single door.

— Mike Schroepfer, Sozzi Unleashed
Key Insight
Automation is what tips a long-standing asymmetry into a present danger: an agent can spend days trying every door and window until one gives, while the defender has to get every door right. Having run large-scale systems himself, he calls this a day-zero risk -- the systems need hardening now.
05The Triage

Not the Sci-Fi Risk

He can describe the runaway-AI scenario but deliberately does not dwell there, putting his focus on the near-term risks and benefits that are already real.

I don't spend a lot of time worrying about the existential risk stuff.

— Mike Schroepfer, Sozzi Unleashed
Key Insight
This is triage, not denial. He can state the runaway-AI scenario -- AI self-replicating and competing with humanity for resources -- yet argues the better focus is the near-term issues and the benefits already arriving, like the companies he backs shipping faster. The prioritization, not the dismissal, is the point.