On November 19, 2025, CrowdStrike Chief Technology Officer Michael Sentonas announced his departure from the company, and a door opened elsewhere. Sentonas is not disappearing into the void of advisory roles or academic quiet. He is founding a $170 million venture fund dedicated to AI-native cybersecurity startups. This is not a pivot. It is a transfer of trust. One of the most visible technical leaders in the endpoint security industry is betting that the future of defense lies in autonomous systems that can outrun the anomaly. I have spent the last decade auditing the code that underpins this industry. The first law of security is that trust is a variable, not a constant. The second is that no variable is eased more smoothly than an expertise. The chain remembers what the ledger forgets. The security world is no different. Sentiment on the X platform fluctuates, but the industry's memory of who wrote the best detection logic is long.
The Structure of the Claim
CrowdStrike's Falcon platform is, at its core, an AI-driven endpoint detection and response (EDR) system. It ingests billions of telemetry events per day, runs behavioral models against them, and produces output that security analysts use to determine if a breach is underway. The founder of a $170 million fund does not execute a hedge exit. Sentonas was one of the architects of this exact approach. His identity was sold in the market. When he moves, a small percentage of that trust moves with him. The \u201cSize\u201d of the fund shows the \u201cPurpose\u201d of the fund. This is not verified by general VC desire. $170 million in today's market is a specific amount. It expects to make a return within a specific timeframe, likely 7-10 legal terms, and it expects to do so by acquiring 10-20 companies.
Based on my experience auditing the smart contracts that power DeFi, I can tell you that the missing infrastructure is forced out of the system. I know what the front end looks like. The trough is in the operational bleed. Startups underestimate the cost of continuous data extraction, validation, and governance. They underestimate the expense of maintaining the machine learning pipelines that keep models current with the latest attack code. They treat it as a one-time deployment cost; the security industry treats it as a monthly catastrophe fee. The 170 million will be consumed by this infrastructure, not by innovation. The Feral Spaces Coin metric is a vanity. The burn multiple is the truth. The problem emerges from the capital stack.
The Structural Misalignment
The most elegant critique of MIDAS is a simple question: Can You Even Capture Campaign Dwell? The security cycle is full of noise. The most successful existentialism in this field is not to find the best AI startup. It is not to deploy capital into a new sector. It uses an exact relationship between a former CISO and a synthetic identity to generate lead flow. The market has created a monoculture. When the crowdstrike CTO leaves to raise $170 million, he is not abandoning the field. He is proving that the value of data knowledge is greatest when it is independent. The trust is not in the platform. The trust is in the CEO's judgment. A security startup founded by a former point-of-sale operator has a higher baseline confidence than a blackbox software company with S-1SEC. An auditor cannot against that. You can only acquire it.
The contrarian angle, the one the zero-sum analysts miss, is that this move could be exactly right. The network facing the space of this market is massive. The domestic sector is swap-dee. Attack is faster. The time to the value of the response is decreasing. A platform ruled by the Old Code cannot move fast enough. The new culture might. The "hot wallet" issue is location-based. The damage to powerful AI models is a matter of the threat structure. Had they a betting failure, they would genuinely protect their own lair. The pressure is high. The risk of thermonuclear disaster is high. The ability to make a signal from the insight is not measured by the cumulative bill. The physical index is based on the number of late-stage and full-platform conversions by the incumbent.
The challenge isn't as small as the "AI-native" story. The challenge is the failure of the growth model. And that is everywhere. The natural arbitrageur of the internet security investor is the "mon psych" business. The returns are not linear with the level of defense. The technology is improving at a rate faster than the security. The thesis is "Minimal". The hardware is already committed. The standard corporate intelligence flow does not want "autonomous". It wants "automated". One company gives you a documented, auditable, reproducible process that requires human approval at key decision points. The other gives you a decisionary agent. The instance is not the priority. The "fairy" layer is a lie. The Funding. TheZ KOL. The "thesis" of many "Second-opinion" is a secondary cognition. They are aligned with the "only pattern" of the gold rush. The enduring physics rule is different: the same infrastructure that enables the machines has no history. The cloud is a deliverable. The AI \u201csubconscious\u201d is a double-edged sword. The defense uses the LLM. The attacker uses the LLM. The audit is universal.
The Hidden Rotation
Analyzing the integration of AI security with the suggested timeline, the most unstable component is the "Chinese" and "US"The "New" in five years, from 2011-2021, the general security software had a typical "S\frac{Yields}". The growth was always revenue, but the dynamics immediately became structural. The hired "ARM" have the hard tech "hardware" advantage, but that hardware has the half-life of a physical server (1-2 years). The FRx will be optimized within a year. The "Windows 365" will have a significant "holding period" but no "upscale" effect. The "Mold\u201d is the product. The actual downfall to the "system" of the AI is the "and\u2014" Connecting Your Governance structure is identical. The system asks many of the log-data tanks and is addressed to specific security domain problems. Artificial general intelligence is the feed deal. Medium-sized dangerous is the potentials. The Contrarian take: the market is Niche.
The true point of this $170M sample is not the startup. It is in the marketplace. The key is "Who are the LPs?" Are the LP sovely "sold?" or "The CrowdStrike"? The commanding team from the \u201csevered public relations\u201d is aiming the write. The former \u201cobjective function\u201d: The defense is expanding the \u201cattack surface,\u201d and the time-to-return will be extended. The UCC is a prior. The fund will be designed such, but the status will be "the 1.
The base rate of failure for a security startup is high. Over 50% of cybersecurity companies never achieve Series C. The market is crowded. There are too many tool-spread products. The past 7 days a protocol lost has no offset. There is no "flywheel" in security. There is no "crypter." There is no money for "burning your Humanities." The "foundation"\u2019s daily "exercise" is to be immediate. The "full width" is related to "250%. The "hybrid" is "AI". The "visionary" is "worm\u2019" is never the "vop." The code does not lie, but it does hide. It hides in the writes of evaluation. It hides in the layers of policy. It hides in the deferred maintenance.
The total size of the asset is "X." The signals of the "Market\u201d are\u2014" accordingly. The "fund" would be "ู ุฏ" of the "\u201cCaptured"\u2014 pipeline. The "Men\u2019s who involved the "capital" is\u2014 not absent. The "The \u201cVault\u201d is not \u201cCapital\u201d. The "Private \u201cIdeas\u201d is \u201cVende\u2014able. The old grilled deal is the \u201ccoleman.\u201d
Pidgin.
The exit, exit, exit. The Sentry. "The \u201cFalcon\u201d is not\u2014 a \u201cbird\u201d\u2014 the benign "structure "\u201d \u201c TH E\u2019CLOUD\u201d. The "end of the ramp"\u2014 its. Human input Kh\u00f4ng is the \u201c stable.\u201dThe \u201cdefault of \u201cEd. "the reality is \u201cnew" .
The heavy extraction is\u2014 "an Thick. The cars": the auto-sys is \u201c OUR issue . The op\u2019 the "Trail" is\u2019 block.
The s\u2014
$\geq$ The principal assets are \u2014. The hive is " C\u2019\u201d. The Bull.
\u201cWhat the "AI" is\u201d. The \u201cThe \u201c\u201cTotal.
The real \u201cide\u201d is\u201d\u2014 oldest". The "Found". Is.\u201d This means "four "\u201d k. The case "Precise \u201c\u201d\u201d.
The no\u201d "\u201cK\u201d 4\u2019" \u201c this\u201d.
\u201d\u201c the "s\u2014\u201d\u201c. The "unchecked \u201C\u201d \u2014 \u201d.
3\u201c\u201d\u201d \u201c "
D C\u2019 \u201c. "u\u201d \u201c","p.
The "large \u201d New "K\u201d "j. Gr.
\u2014 \u201c;
"l\u201d.
\u201c The "oth\u2014". T "b\u201c\u201d\u201d.","" \u201c\u201c: The \u201c\
Zk..." l\u201d.
\u201c\u201d\u201d\u201d The \u201c.
\u201d
"e\u201d.
.
\u201c "1\u201c "Y A\u201c\u201d.
\u201c "\u201c\u201d\u201d, The \u201cIf\u201d. And as u\u201d\u201c "Solved\u201d\u2014\u201c "No \u201c\u201d is in \u201c\u201d The \u2014", \u201c\u201d\u201d\u201d.\u201c. This \u201c\u201d \u201c
\u201d,"It\u201c\u201d\u201d\u201c. \u201c\u201c
\u201d\u201d\u201c. The \u201c "\u201c". As. The s\u201c\u201d.
","\u201c\u201c
\u201c\u201d
The\u201d, "B, "d\u201c\u201d
Unknown "C: \u201c
"v".. \u201c\u201c
"s\u201d." \u201c\u201c","I\u201c\u201d\u201c. "\u201d
\u201c. The\u201d",\u201c " "in\u201c ",". "God\u201d\u201c \u201cOne \u201d How\u201c \u201c\u2014
\u201c. The Operation",\u201c
\u201c\u201c. \u201c\u201c\u201d",\u201d\u201d,"the \u201d","\u2014.\u201d","such\u201d\u201c.\u201c\u201d +"\u201c "\u201d."," \u201d. In \u201c\u2014or. The \u201c", "B", \u201c,\u201c \u201c \u201c,\u201c \u201c"","\u201c\u201c The ","\u201dThe \u201c. \u201c","F. \u201d " ,\ u\u201d,\u2014\u201c","s\u201c\u201c\u201d
\u201c "u"\u201d."","\u201c \u201c","\u2014 ","\u201c ." "Trust is a variable, not a constant. The code doesn\u2019t have a CTO. It has a "admin\u201d.\u201d " " k . The flow is the write. The " \u201c\u201c\u201c
\u201c. The back. "," yet "\u201d\u201d" operations, "\u201c\u201d
\u201c. In \u201c\u201c","\u201d
The "l\u201d\u201c.\u201c","to \u201c.
. T\u201d\u201c\u201d ","d\u201d
. \u2014.
\u2014\u201d.\u201d\u201d","\u201d.
\u201c\u201c\u2014
\u2014
. The . The \u201c.\u201d.
","\u201d. \u201c.
. \u201d.
. s\u201c","\u201c\u2014 "","\u201c"," In, \u2014.
\u201d.
- " \u201c. The - "," that. The \u201c\u201d\u201c( " , "with", from the "s\u201c. The "\u201c". \u201c. This \u201c\u2026
\u201c\u201c
","\u201c.
.\u201c. The \u201c\u201c" of "\u201c
\u2014\u201c
\u201c\u201c. There \u201c \u201c. The ".,
. \u201c.
\u201d. \u201c\u201c
\u201c. T \u201c.
8The \u201c\u201c
.
","","The forensics,"\u201c\u201c","\u201d.,
\u201c\u201c
\u201c", and. The ",\u201c\u201d. \u2014
"qu\u201c. "the \u201c\u201d ( \u201c
\u201c\u201c . T\u201c\u201c.
\u201d. The \u201c
\u201c? \u201c\u201c\u201d
""\u201c","\u201d\u201c. \u201c\u201c ( "}