This Week in Startups

Is Chalk.ai the ‘Next Databricks’? + Tollbit’s Bot Paywall for AI Agents | E2167

91 snips
Aug 20, 2025
Guest
Toshit Panigrahi
Guest
Marc Freed-Finnegan
Marc Freed-Finnegan, Co-founder and CEO of Chalk, discusses tackling data freshness in AI by providing real-time data pipelines that boost inference speed, positioning Chalk as a potential leader in the field. Toshit Panigrahi, CEO of Tollbit, shares insights on their innovative 'bot paywall' designed to change how AI agents access content, signaling a shift in data licensing trends. Listen in as they dive into the wrap-up on the evolving landscape of AI monetization and data accessibility, hinting at what the future holds for this dynamic industry.
Ask episode
AI Snips
Chapters
Transcript
Episode notes
INSIGHT

Inference Compute Is The New Bottleneck

  • Inference compute is rapidly outgrowing training compute as models are run continuously in production.
  • Fresh, real-time data at inference time transforms user experiences and drives much of AI's current compute demand.
ANECDOTE

Whatnot Example: From Nightly To Instant

  • Whatnot used nightly recommendations and missed same-day browsing signals until Chalk enabled instant updates.
  • Chalk fetched source data at inference time so recommendations reflected the user's immediate activity.
INSIGHT

Batch Pipelines Leave Freshness Gap

  • Traditional batch systems (Databricks/Snowflake) produce stale cached features for low-latency serving.
  • The tension: fast serving vs fresh data creates a market for real-time pipelines between feature stores and batch training.
Get the Snipd Podcast app to discover more snips from this episode
Get the app