Benchmark Partner Jack Altman: No One Truly Understands This Round of AI
Benchmark partner Jack Altman stated that the most important point in understanding the current cycle of artificial intelligence is that no one really knows what they are talking about. Even very smart and well-informed individuals continuously overturn their judgments, only to reverse them again in three months.
He did not name specific individuals or instances of reversal. Altman founded the HR software company Lattice, later operated Alt Capital, and is now a partner at Benchmark while hosting the interview podcast Uncapped. His brother, Sam Altman, leads OpenAI. He himself wrote in January that as one gets older, they become less confident in a single viewpoint; he has quoted Vinod Khosla's prediction that "80% of economically valuable work will be done by AI within five years" and relayed concerns from Founders Fund about valuations detaching from fundamentals, similar to the unrest seen in 2021. This round of capital expenditure, model capabilities, coding agents, and data center robots are all accelerating simultaneously, and public predictions on "job replacement timelines, whether open source can catch up to closed source, and whether there are barriers at the application layer" have been repeatedly revised, which is the industry context for his judgment rather than a list of cases he provided.
The remarks did not present new data or new investments, which discounts the credibility of the prediction market and memorandums. A reversal every three months means that the narrative of the same group of "insiders" cannot be considered stable.
Mechanically, this is a cognitive discount, not an asset sell-off. The buyers are early funds that want to write uncertainty as options, while the sellers treat a single roadmap as certainty. Funds continue to flow into computing power and applications, but terms will be shorter and milestones more fragmented. The beneficiaries are founders who can rewrite products between reversals, while the pressured parties are LPs who treat last quarter's memorandum as a three-year strategy. Benchmark's tradition is to focus on a few high-quality investments, and this tweet serves as a reminder: the premise of concentrated investment—understanding the cycle—does not hold this time.
In terms of supplementary statements, he only makes judgments without providing names, transforming the authority of the "smart and insider" into samples that can also reverse their statements.
Source: Public information
ABAB AI Insight
Jack Altman dismantles the authority of insiders: being in the room does not equate to providing a timeline. A reversal every three months indicates that the model's capabilities are evolving faster than the rewriting cycle of investment memorandums. His background at Lattice makes him accustomed to organizing verifiable metrics, but the AI cycle lacks hard indicators of the same frequency, leaving only evaluation rankings and financing news, making opinions as variable as the weather. With his brother in a modeling company and himself in an early-stage fund, this tweet also serves as a warning to LPs: do not mistake the family name for internal consensus.
The capital path is characterized by high belief encountering high revision rates. Venture capital relies on stories to lock in ten-year funds, while model labs adjust capability boundaries quarterly. The clocks of the two do not sync, leading to instances of "very smart people publicly reversing their statements." Khosla's prediction of 80% job displacement and the valuation unrest of 2021 can coexist on the same podcast timeline, indicating that both bullish narratives and bubble narratives share the same guests. Money continues to flow into data centers because the computing power bills are more concrete than opinions; equity stories must shorten according to the frequency of reversals.
Analogies can be drawn to the early mobile internet's repeated concerns over whether "applications would be consumed by operating systems" and the back-and-forth during the crypto cycle regarding whether "the infrastructure is sufficient, and it's time to build applications." The industry is at a stage where capital expenditure has landed, but application profits have yet to stabilize. Just because no one can clarify does not mean no one is betting; it only means that the validity period of the betting explanations is measured quarterly.
Structural changes indicate a transfer of pricing power. The explanation power shifts from "insider consensus" to verifiable revenue and retention. The mechanism is that capability leaps outpace due diligence templates, shortening the half-life of authoritative statements; the cost of reversal is lower than admitting ignorance; thus, what is traded in the public market is the next memorandum, not the previous truth. The true information of the cycle is that reversals themselves have become the norm.
ABAB News · Cognitive Law
The authority of insiders often has a validity period shorter than a quarter.
The smarter and more insider one is, the more likely they are to publicly reverse their statements in three months.
In a cycle where no one can clarify, money continues to flow according to computing power bills.