• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
  • HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    AI

    Santiago Valdarrama Reports Tripled Token Costs

    ML engineer Santiago Valdarrama reported sudden increases in token expenses from his agents.

    SA
    2 Sources, 29d ago, first seen 29d ago

    TLDR

    Santiago Valdarrama, a computer scientist and ML engineer who teaches production AI work, posted that his token costs tripled over a few weeks. He stated he ran only a few agents and found nothing broken in the setup. The claim comes directly from his own account of billing changes while the configuration stayed the same. No further details on the exact agents or providers appear in the post.

    Combined views

    14.5K

    2 Sources, first seen 29d ago

    Combined views

    14.5K

    2 Sources, first seen 29d ago

    34 likes
    34 likes
    9 comments
    42 saves
    6 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    9 comments
    42 saves
    6 reposts

    Sentiment

    Positive——Negative

    Summary

    Not enough discussion yet.

    No sentiment analysis available yet.

    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet

    2 Sources

    @svpinoOur token costs tripled over a few weeks. We only had a few agents running, and nothing was broken, but costs went 3x what we expected. We looked at the traces, and this is what we found: • Agents were reading more files • They were loading more context • They were creating more detailed plans • They were verifying their work more frequently • They were troubleshooting mistakes • They were retrying failed steps Nothing here is broken. This is exactly what we wanted. But, as models improve, so do our agents and their ability to spend more time on a task. Here is the important lesson: 99% of the teams I collaborate with would go crazy because they would have no clue why their bill tripled. This is why observability matters so much. Observability is the only way we can understand what our systems are doing, especially as they become increasingly autonomous, probabilistic, and expensive. The @honeycombio Team & Liz Fong-Jones are running a free observability masterclass with six sessions, covering everything from instrumentation fundamentals to the business case and cost. The next class is live on September 2 for engineering leaders. You can register here and catch up on the previous sessions: https://fandf.co/4xUUn9T Thanks to the team for partnering with me on this post.

    2 Sources

    @svpinoOur token costs tripled over a few weeks. We only had a few agents running, and nothing was broken, but costs went 3x what we expected. We looked at the traces, and this is what we found: • Agents were reading more files • They were loading more context • They were creating more detailed plans • They were verifying their work more frequently • They were troubleshooting mistakes • They were retrying failed steps Nothing here is broken. This is exactly what we wanted. But, as models improve, so do our agents and their ability to spend more time on a task. Here is the important lesson: 99% of the teams I collaborate with would go crazy because they would have no clue why their bill tripled. This is why observability matters so much. Observability is the only way we can understand what our systems are doing, especially as they become increasingly autonomous, probabilistic, and expensive. The @honeycombio Team & Liz Fong-Jones are running a free observability masterclass with six sessions, covering everything from instrumentation fundamentals to the business case and cost. The next class is live on September 2 for engineering leaders. You can register here and catch up on the previous sessions: https://fandf.co/4xUUn9T Thanks to the team for partnering with me on this post.