Do you know a guy? | 99d Summary

1. Key Themes

Theme: Agents erase the "laziness subsidy" that funds many consumer deals

Most economic "deals" are funded by user indifference

"In normal life (pre agents) not everyone is a ruthless optimizer. Most of the time, most people are content to take the path of least resistance while a few freaks value-max their way through life. Essentially your laziness or indifference pays for my margin or surplus."

Free optimization makes the first-order "everyone wins" conclusion wrong

"The obvious - and wrong - first order conclusion is that everyone will be more efficient and find lots of new value. But this is a relative value game; not everyone can cleverly cut the line, get special attention, or enjoy negative unit economics/subsidies. No system can support a population of 100% ruthless optimizers and rule exploiters while maintaining its steady state."

The second-order effect is that subsidies disappear rather than spread

"So the second order (correct) effect is that rather than being widely enjoyed, a bunch of these subsidies and deals will go away entirely."

Examples cited across consumer and financial products:

"Credit card rewards are the most obvious example of this. The disinterested subsidize the reward hackers." "Subscription businesses often count on low utilization and cancellation rates which keeps the prices lower, subsidizing the optimizers and value extractors." "Customer service and retention offers work the same way. The cable discount exists because few people call and threaten to cancel."

Theme: Agent-vs-agent dynamics raise system-wide costs and barriers

Financial systems are priced on assumptions of low agency

"Mortgage rates are priced on the assumption only a limited number of people will refinance when rates drop. Morgan Stanley thinks AI agents could push refi frequency 2x, leading to basis point increases in mortgage rates overall." "Banks rely on cheap (low interest) deposits to make loans. Apollo speculated that agents could cause banking runs and drive up the cost of capital economy-wide by sweeping cash automatically."

Healthcare shows the arms-race dynamic already in motion

"In medical billing, my relative unwillingness/inability to upcode makes it profitable for you to do so. Now it's ubiquitous, driving up the overall administrative costs according to BCBS as both sides deploy AI against each other."

Agents create demand, not supply, for scarce assets

"The introduction/proliferation of the infinite optimization machine (on-demand intelligence and personal intelligence) won't create new supply of scarce assets (physical resources, attention, etc.) but will substantially/infinitely create new demand for them, especially as agents recursively fight with one another."

Defensive countermeasures will erode deals

"Deals will go away because of active defense as agents are deployed against agents for the purpose of obfuscation and obstruction."

Theme: Edge migrates from agency to proprietary, human-network channels

Rote agent execution can't replicate relationships or physical presence

"Edge will move from 'agency' (really, persistence and optimization - and at the limit a willingness to trash the commons) to more proprietary channels that rote agent execution can't replicate. In the simplest terms that means 'knowing a guy' or 'showing up in a room.'"

The end state is a barbell economy

"The whole economy will look like HFT on one hand and 'knowing a guy' on the other."

Legible hacks lose their value

"So that probably means legible 'hacks' go away as this meritocratic subsidy."

Theme: Ubiquitous agents are redistributive or zero-sum, not broadly value-creating

Equal capabilities cancel out individual advantage

"But everyone will have them with the same rough capabilities, which clearly is redistributive at best and value destructive/zero sum at worst."

Human attention remains the binding constraint

"And even where compute sufficient to execute long run or complex tasks is wildly inequitable, human attention will remain the primary bottleneck on real world actions and scarce resources. Again, redistributive."

2. Contrarian Perspectives

Agents will make barriers to surplus higher, not lower

The consensus view is that AI agents democratize optimization and lift everyone. Rechtman argues the opposite: because deals depend on most people not optimizing, universal optimization destroys the deals.

"In the near future of ubiquitous personal agents, the economic SOP will shift toward ruthless optimization and you should expect higher, not lower barriers to surplus and service everywhere."

His evidence is that many systems explicitly price in low user agency (mortgage refinancing frequency, bank deposit stickiness, subscription churn, retention offers).

"The list is basically infinite. Dating apps, driving directions, networking, the court system - so much of our economy and society relies on low agency (low propensity for ruthless exploitation and optimization) to function."

Agent-driven "meritocracy" is partly illusory

While agents may help in some arenas, the net effect is not broad gains.

"In some arenas agents will make life more meritocratic, because a person can do more things more easily if they are willing to."

But because capability is widely shared, the gains wash out, and the scarce resource (attention) cannot be expanded by compute.

AI investing is starting to abandon reputation norms

Venture's iterated game normally keeps behavior in check, but big dollars and profit-taking moments break the norm.

"But again and again, when the dollars get big enough and/or it comes time to take profits, those norms go out the window. When it really might be 'one last ride' who cares what they say about you after it's done?" "That is clearly starting to happen in AI investing. The only surprise is that it took this long."

3. Companies Identified

Morgan Stanley

Description: Global investment bank
Why mentioned: Cited as the source of a forecast on how agents could change mortgage dynamics
Quote: "Morgan Stanley thinks AI agents could push refi frequency 2x, leading to basis point increases in mortgage rates overall."

Apollo

Description: Alternative asset manager
Why mentioned: Speculated that agents could trigger bank runs and raise the cost of capital
Quote: "Apollo speculated that agents could cause banking runs and drive up the cost of capital economy-wide by sweeping cash automatically."

BCBS (Blue Cross Blue Shield)

Description: Health insurance association
Why mentioned: Source for evidence that AI-driven upcoding is raising administrative costs
Quote: "Now it's ubiquitous, driving up the overall administrative costs according to BCBS as both sides deploy AI against each other."

Slow Ventures

Description: Venture capital firm (author's employer)
Why mentioned: Author's affiliation and investing context
Quote: "I'm a partner at Slow Ventures, where I lead pre/seed rounds from a ≈$325M fund."

4. People Identified

Yoni Rechtman

Description: Partner at Slow Ventures; author of 99d
Why mentioned: Author of the piece
Quote: "I'm a generalist investor looking for weird takes on important stories: N-of-1 companies taking non-obvious approaches to markets that matter."

Hunter Walk

Description: Investor and writer (linked in "Elsewhere")
Why mentioned: His linked post is about AI slop overwhelming his inbox, which parallels the article's agent-driven degradation theme
Quote: "I used to respond to every cold email but now AI slop is killing my inbox | Hunter Walk"

5. Operating Insights

Build moats that agents can't replicate

Competitive advantage will shift away from rote persistence and toward relationships and physical presence.

"Edge will move from 'agency'... to more proprietary channels that rote agent execution can't replicate. In the simplest terms that means 'knowing a guy' or 'showing up in a room.'"

Stress-test any business model that depends on customer inertia

If your margins rely on low utilization, low cancellation, or low refinancing, assume agents will push those behaviors toward the optimized extreme.

"Subscription businesses often count on low utilization and cancellation rates which keeps the prices lower, subsidizing the optimizers and value extractors."

Expect to need an agent defensively, and expect counterparties to deploy them too

Optimization becomes table stakes rather than an edge.

"The whole economy is gonna become more competitive and everyone will NEED to become an optimizer (or 'hire' an agent to do so) to overcome the newly erected barriers to surplus value."

6. Overlooked Insights

Reputation as the honesty mechanism in venture is fragile at big-dollar moments

Beyond the AI-investing observation, the structural point is that long-term reputation costs only constrain behavior when future interactions matter, which is exactly when profit-taking weakens them.

"So everyone has to manage their reputation for the sake of long term profits, sometimes to the detriment of maximizing their short term P&L."

Agent proliferation implies a tragedy-of-the-commons dynamic

The article notes in passing that the limit of optimization includes degrading shared resources.

"...at the limit a willingness to trash the commons"

This hints at a broader class of risks (spam, reservation DDOS, inbox flooding) where individual agent optimization degrades shared systems.