Why No Vendor Will Tell the Boardroom That 100x AI Productivity Is Fiction

The 100x productivity pitch is sold to executives whose careers depend on believing it, by vendors whose contracts depend on those executives never being contradicted. That is the structural problem Nik Suresh names in a 19 July 2026 essay, and it is the part the discourse around "AI transformation" rarely acknowledges directly. Suresh is a consultant who has run point on roughly 300 sales conversations and technical engagements over the life of his blog. The pattern he reports from that vantage point is not "AI doesn't work." It is that the people choosing, buying, and publicly defending AI strategy inside large organizations are almost never the people in a position to honestly evaluate it. (Ludicity)

Public companies are announcing AI productivity gains after purchasing Copilot licenses and "declaring victory," Suresh reports, while internal chatbots launch without tracking basic usage metrics, or track metrics employees can game. He cites a Mitsubishi voice bot that signed up a customer in late 2024 and never called back over the following six months. Project leaders avoid the basic usage data, or track easily gamed numbers. None of these are capability stories. They are stories about who inside the organization is willing to say "this isn't working" without being removed, and who is structurally unable to say it. (Ludicity)

Suresh's load-bearing anecdotes are anonymized. One executive who had never personally used ChatGPT or any other AI tool authored an AI-centered technical strategy for a $2 billion-plus revenue organization, per his account. An engineer at a separate company, watching colleagues check a token-based productivity leaderboard, parallel-checked out a Go repository so AI could rewrite it in Zig, presenting the rewrite as a job-protection move. Both match what readers suspect; neither is independently verifiable in the source. (Ludicity)

Hermit Tech, Suresh's firm, markets itself as rescuing failing software projects using what it calls "ancient techniques" drawn from 1986 to 1999 software engineering literature. On Hacker News, commenters led by user A1kmm pointed out that this positions the firm as one whose sample of "AI projects" is, by construction, the projects that have already failed. The 0% success rate Suresh reports on AI projects over 1.5 years is therefore best read as a reported observation from a particular vantage point, not an industry statistic. Suresh's own epistemic note, that his wellbeing is not contingent on paying lip service to the AI trend, suggests he is aware of the position he is writing from. (Hacker News)

If Hermit Tech only sees the projects that are already failing, the people inside organizations who are actually trying to make AI work productively are not in Suresh's data at all. That makes the piece a map of which projects end up at a firm whose pitch is "we will fix what AI broke," not an industry verdict on AI itself. The structural argument survives the selection-bias critique; the specific failure-rate number does not. (Hermit Tech)

Suresh reports that vendor executives cannot honestly say 100x productivity claims are implausible, because doing so would undermine the customer executives who authorized the contract and put the vendor's own relationship at risk. Mitchell Hashimoto, co-founder of HashiCorp and lead on the Ghostty terminal, put a sharper version of the same point on X: "I strongly believe there are entire companies right now under heavy AI psychosis and it's impossible to have rational conversations with them about it." That post is embedded in Suresh's essay; the original tweet has not been independently verified for this piece, so the quote should be treated as a caveated attribution rather than a hard quote on the record. (Ludicity)

The second-order cost is the strategy that does not get written while the AI strategy consumes executive attention, the rebuild that does not get staffed while the AI rewrite consumes engineering cycles, the project that does not get funded because it does not survive an AI-themed board narrative. Suresh calls AI strategy a credibility and political artifact inside organizations, not a build-versus-buy calculation. The constructive exit ramp in the piece is reporting that names the gap instead of applauding the press release, internal accountability for AI claims, and vendor pricing that does not depend on customer executives never being contradicted. (Simon Willison)

Q3 board cycles and Q4 enterprise renewals are when this lock-in will be most legible from outside the buyer-seller relationship. The question worth carrying into the next AI strategy meeting is who in the room is paid to call the 100x claim a fiction, and what happens to them when they do. (Ludicity)