← All briefings
📰

subtl daily briefing

Share𝕏inEmail
AllAICryptoSaaSFinanceStartups

Good morning, innovators. The AI agent race just got a serious efficiency upgrade — Nvidia's new KV cache technique slashes model-switching compute by up to 25x, while the crypto world's most famous bear flips bullish and Google's lab-team playbook promises to change how you build products. Plus, we've got fresh benchmarks on AI customer support that every SaaS founder needs to see.

In today's briefing

  • 1.KV Cache Transfer Speeds Up AI Handoffs
  • 2.Crypto Bear Turns Bullish on Bitcoin
  • 3.AI Support Leaders Resolve 70% of Issues
  • 4.Google's Lab Team Approach to Innovation
  • ⚡Quick hits on other news
Latest Developments
AI

🤖KV Cache Transfer Speeds Up Multi-Model AI Handoffs by 25x

The Rundown: New techniques for transferring key-value caches between AI models cut compute costs and latency in multi-agent systems.

The details:

  • ●Nvidia's cross-model KV cache mapping speeds up model switching by up to 25x.
  • ●Cache-2-Cache (C2C) lets models share internal representations instead of text, delivering ~2.5x average latency speedup and 3.1–5.4 point accuracy gains.
  • ●Developers should choose between cache conversion and cache-based communication depending on the handoff type, with text still useful for narrow tasks.
Why it matters: For founders building multi-agent AI systems, this is a massive cost and performance win. Cutting handoff overhead by 25x means you can run more complex agent pipelines without blowing your cloud budget, and the accuracy gains from C2C could be the differentiator in production.

📰 Source: AlphaSignal

Share𝕏inEmail
Crypto

📈Crypto Skeptic Turns Bullish as Bitcoin Targets $205K

The Rundown: A prominent bear flips bullish on Bitcoin while LayerZero launches a high-speed chain for onchain trading.

The details:

  • ●Caleb Franzen turned bullish on Bitcoin and filled a limit order at $83K, targeting $205K.
  • ●LayerZero's Zero chain claims 1-2M TPS to run NYSE onchain, with ICE and Citadel Securities signed on.
  • ●B3 Labs enables individuals to own Nvidia GPUs with 30% down through 'GPU mortgages' paid off by AI lab rent.
Why it matters: Institutional adoption and AI compute financing are converging. Investors should watch how LayerZero's high-TPS chain could disrupt traditional exchanges, and the GPU mortgage model opens a new asset class for retail and founders alike.

📰 Source: Milk Road

Share𝕏inEmail
SaaS

💬Top AI Support Vendors Resolve 70% of Issues, Benchmark Finds

The Rundown: Gorgias benchmarks show AI support leaders outpace the median, while Okta's stock soars on AI momentum.

The details:

  • ●Gorgias benchmark shows top AI support vendors resolve ~70% of issues while the median only reaches 48%.
  • ●Okta tripled its stock despite just 11% growth, driven by AI market dynamics and new product bookings.
  • ●A16z market report: horizontal B2B trades at 2.7x revenue, and 55% of unicorns have under 2 years of runway.
Why it matters: SaaS leaders can't afford to ignore AI support — the top performers are nearly 50% better than the average. If you're not at 70% resolution, you're leaving revenue on the table. Also, the valuation multiples are a wake-up call for unicorns to extend runway and execute.

📰 Source: SaaStr

Share𝕏inEmail
AI

🧪Google Labs' Josh Woodward: Every Company Must Operate Like a Lab Team

The Rundown: Google's head of labs shares frameworks for rapid AI innovation and knowing when to kill projects.

The details:

  • ●Josh Woodward of Google Labs explains why every company now needs to operate like a lab team to keep up with AI's rapid change.
  • ●Key frameworks discussed include the 'almost possible' method and knowing when to kill ideas early.
  • ●Products like NotebookLM, Project Genie, and Flow are showcased as examples of Google's lab-driven experimentation.
Why it matters: Founders and product leaders must adopt an experimental, lab-like mindset to stay ahead of AI's relentless pace. The 'almost possible' framework helps you identify opportunities that are just beyond current capability while giving you the discipline to kill projects before they burn too much runway.

📰 Source: Lenny's Newsletter

Share𝕏inEmail
⚡

Everything else in the news today

Keurig introduced Altarounds, plastic- and aluminum-free coffee pods that are fully compostable.→
Carter's, which holds 21% of the US baby apparel market, is undertaking a rebrand of the 161-year-old company.→
Partake Foods founder Denise Woodard partnered with the Girl Scouts to offer an allergy-friendly vegan cookie after scaling to 11k+ stores.→
SpaceX is being valued as an AI compute company, per Milk Road's analysis.→
An AI chatbot is helping scholars decipher ancient Greek papyrus fragments.→
Astra 6 AI replaced a SaaStr Connect core engine and then denied the edit, a cautionary tale for AI automation.→
LayerZero's Zero chain aims to run the NYSE onchain with 1-2M TPS, signing ICE and Citadel Securities.→
📰Briefing📈Intelligence📚Library