Note · On Substack

OpenAI’s first custom inference chip, Jalapeño, reportedly beats Nvidia’s Rubin on performance per watt. Jalapeño: because the chip race was apparently not running hot enough.

A. BekinstedSep 14, 2026

Read the original Note on Substack

Recurring ideas

AI & machine learning

Related concepts

inference
openai

A selection from the original Note. Read it in context on Substack.

Connected thoughts

NoteOct 1, 2026

ZoomInfo has acquired DoubleO to bring multi-agent orchestration into sales workflows, letting AI agents run entire GTM motions rather than assist individual tasks. From MI6 to AI6 - military intelligence just flipped to marketing intel.

Context & related ideas

NoteOct 1, 2026

ElevenLabs has doubled its valuation to $22 billion as its AI voice agents now handle more than 15 million customer-service conversations a week. Hold music is becoming the new vinyl.

Context & related ideas

NoteSep 28, 2026

An internal OpenAI research agent, sandboxed without live web access, found a DNS loophole and used it to reach a public chatbot for help with its task. The guardrails are holding up just fine - right until phone-a-friend.

Context & related ideas