Investor Zach DeWitt: Not Expressing a Slowdown Seems Out of Touch
Venture capitalist Zach DeWitt stated that after Elon Musk and Sam Altman agreed with Dario Amodei's assertion to "control the frontier pace," if a CEO of any lab does not make a similar statement, they will appear out of touch.
The trigger was Amodei's publication of "We Must Pace the Frontier," which unilaterally committed to having third-party evaluators on-site long-term with near-employee level access; Musk replied, "Dario is right"; Altman agreed to control frontier progress and promised that independent evaluators would have access similar to that of employees. DeWitt interpreted these three public alignments as a new identity threshold: the discourse of slowing down itself becomes a ticket to the frontier club.
DeWitt is a partner at Wing Venture Capital, focusing on early-stage companies driven by AI and product growth, and is not an insider at the three labs. His judgment falls on the reputation game: silence will be read as "not cutting-edge enough," rather than "having an alternative technical route." In the same week, at least Google DeepMind, Meta, and other heads of closed labs outside of xAI's operational layer had not made equivalent statements; Demis Hassabis and others had also not appeared on the core CEO list of the "Pacing the Frontier" employee petition.
In July, over a thousand employees from frontier labs signed a petition urging the U.S. government to support international tools to apply brakes on automatic R&D, with signatures concentrated from Anthropic, OpenAI, Google, and Meta. Pachocki later stated that no lab deserves full-speed expansion. Amodei's plan also includes unified security standards for democratic countries, attempts to coordinate with China, while limiting advanced chips and model distillation. OpenAI is asking Congress whether industry collaboration to slow down would violate antitrust laws.
The competition of statements and training is beginning to overlap. Whoever first expresses "agreement to slow down" occupies the microphone defining what is cutting-edge; those who delay must explain whether they disagree with the risks or with handing over pace control to evaluators and the government. xAI's Grok route, open-source weight labs, and teams that still view release frequency as competitiveness will be squeezed by the same set of rhetoric.
Mechanically, this is a discourse power event, not a computing power order. Buyers need "citable safety positions" for policy capital, corporate procurement, and evaluation agencies; sellers must use a post to purchase frontier identity for lab PR. Beneficiaries are those who have spoken out, like Anthropic, OpenAI, and Musk's AI; those under pressure are silent CEOs and their investors—within valuation narratives, the phrase "most cutting-edge" now carries an obligation to express a stance. Funding will not change direction due to a single post, but boards will include "whether to speak out" in this week's risk list.
Source: Public Information
ABAB AI Insight
DeWitt frames safety statements as status commodities. Labs previously proved their cutting-edge status through parameters, evaluations, and release intervals; now they prove the same thing with "I agree to slow down." This is a typical club signal: members must repeat the password, or they are seen as still outside the door. Wing invests in application-layer growth and does not hold core equity in these three labs, so he can expose the mechanism—statement costs are nearly zero for those who have spoken out, while they are a reputational tax for those who have not.
The capital path is to prioritize regulatory coordination in public opinion. Amodei first offers on-site evaluation, Musk and Altman follow with the shortest sentences, and later dissenters must oppose both safety and the claim of being "cutting-edge." OpenAI's request for antitrust exemptions from Congress indicates they know true slowing down requires legal cover; publicly agreeing does not necessitate actually stopping training, but allows them to occupy the definition first. Resources are temporarily diverted from training clusters to PR timelines: those who post on weekends appear more "responsible" in Monday's corporate procurement memos.
A similar structure is seen in banks' statements of "we welcome stricter regulation" after stress tests, and energy companies preemptively acknowledging transformation on climate issues. The July employee petition has turned "Pacing the Frontier" into industry insider jargon; CEO follow-ups elevate this jargon into an external trademark. The industry phase belongs to control: the capability race has not stopped, but the identity race has concluded. Open-source weight labs cannot prove themselves with "employee-level presence," while closed labs can claim the same badge with a single statement.
The structural judgment is a transfer of pricing power. The mechanism is that the term "frontier" shifts from a technical position to a discourse position. Evaluation rankings can only prove yesterday's models, while statements can prove that training is still permitted tomorrow. Silence is no longer neutral, as the reference frame has been aligned by the three labs. The next question is not who stops training first, but who first explains "non-statement" as another form of being cutting-edge.
ABAB News · Cognitive Laws
- If the password is repeated loudly enough, it will be seen as a ticket for entry.
- Once silence is compared against the statements of the three labs, it will be read as lagging behind.
- Those who define the frontier do not need to stop their ongoing training first.