The tweet was deleted by the author.
But we saved everything 🙂.
Tomasz Tunguz reported burning 84 million tokens on February 28. He explained that the tokens were consumed while researching companies, drafting memos, and running agents.
Tunguz noted that running Kimi K2.5, a serverless model accessed via API, would have resulted in a $756 expenditure for a single day's work if billed at Claude or OpenAI's blended rate of $9 per million tokens.
Tunguz’s analysis of operational costs in AI deployments adds to an ongoing conversation about efficiency and scale within technology and finance. His recent assessment of the substantial expenses tied to token usage recalls earlier scrutiny applied to high-profile events such as the $29.4 billion Saudi Aramco IPO, where the implications of record-breaking financial moves were meticulously examined. Similarly, his in-depth look at organizational optimization found in the Block headcount reduction underscores his continued focus on how strategic decisions impact market dynamics and operational sustainability.