AI Projects - October 9, 2026

Week of: October 9, 2026

Overview

The week's most concrete applied-AI work came from individual builders and vendor education teams: Simon Willison shipped both a voice-based feature for his own blog and the 1.0 release of token-counting tool ttok, while Zapier and HubSpot published practical implementation guides — Zapier on rolling out AI adoption across teams, and HubSpot on answer engine optimization (AEO) across a dozen marketing roles. On the startup side, Latent Space interviewed Periodic Labs' Liam Fedus and Ekin Dogus Cubuk about synthesis and materials work, and its AINews desk noted a quiet news day overall.

Stories

1. Simon Willison ships a blog feature built using his voice

Source: Simon Willison Link: https://simonwillison.net/2026/Oct/9/built-using-my-voice/

Willison published a new feature for his personal blog, describing it as "built using my voice." The post is a first-party writeup of a self-directed project rather than a product announcement; see source for the implementation details and how voice was used in the build.

It's a small but representative example of the pattern this beat tracks: individual practitioners building and documenting their own AI-assisted tooling on top of personal infrastructure, then publishing the result as a case study others can copy.

Impact Analysis: Worth reading as a template for how one developer wires AI into a live site rather than a demo.

2. ttok reaches 1.0, a day after 0.4

Source: Simon Willison Link: https://simonwillison.net/2026/Oct/9/ttok/

Willison released ttok 1.0, following a 0.4 release the previous day. The two posts bracket a rapid iteration cycle on the same tool; see source for what changed between versions and what the 1.0 milestone covers.

Token-counting utilities sit underneath a lot of practical AI work — prompt budgeting, cost estimation, and chunking decisions in retrieval pipelines all depend on knowing how much text a model will actually consume. A stable 1.0 makes that dependency easier to adopt in automation scripts where version churn is a liability.

Impact Analysis: Any team estimating per-call costs or fitting prompts into context windows should evaluate a pinned, versioned token counter.

3. Zapier publishes a practical guide to AI adoption

Source: Zapier Link: https://zapier.com/blog/ai-adoption

Zapier published an AI adoption guide framed around a familiar tension: enthusiasm for new tools versus fatigue with them. The piece notes AI is already helping teams save time and cut repetitive work, while introducing new problems including messy rollouts, tool overload, unclear guidelines, and other real challenges.

For automation builders, this is the vendor-side articulation of a demand signal — organizations are adopting AI faster than their internal guidelines and rollout processes can absorb. That gap is where implementation consulting, internal tooling, and governance work gets sold.

Impact Analysis: Adoption friction, not model capability, is the near-term bottleneck for teams deploying AI across existing workflows.

4. HubSpot targets answer engine optimization with role-by-role guides

Source: HubSpot Link: https://blog.hubspot.com/marketing/aeo-guide-for-content-strategists

HubSpot published an AEO guide aimed at content strategists, framing the core question as: when buyers ask ChatGPT, Gemini, or Perplexity about a category, which pages actually get cited? The piece builds on the assumption that content teams already know how to build for search, but citation in answer engines is a different problem. It sits within a broader set of AEO guides the company published the same day for other roles.

AEO is emerging as a distinct workstream with real budget behind it, and it maps almost directly onto automation: monitoring what answer engines say about a brand, tracking citations across many queries, and feeding structured content into those systems are all pipeline-shaped tasks.

Impact Analysis: AI visibility is becoming a measurable marketing function, and the tooling to track it is still early.

5. Latent Space interviews Periodic Labs on AI for synthesis and materials

Source: Latent Space Link: https://www.latent.space/p/periodic

Latent Space published an episode with Periodic Labs' Liam Fedus and Ekin Dogus Cubuk, billed as a crossover between its Science and Engineering shows, covering the path "from semiconductors to superconductors," with a forward-deployed engineering angle. See source for the technical discussion.

Periodic Labs is the kind of applied-AI company this beat watches: AI applied to physical-world research rather than to software alone, with a forward-deployed engineering motion that implies bespoke customer work alongside the research. That hybrid model is worth tracking as a template for deep-tech AI startups.

Impact Analysis: Keep an eye on AI-for-science startups selling forward-deployed engineering alongside research, not just models.

Source Links

• Simon Willison - A new feature for my blog, built using my voice
• Simon Willison - ttok 1.0
• Zapier - AI adoption: A practical guide from Zapier
• HubSpot - AEO for content strategists: How to move the needle on citations
• Latent Space - Synthesis Superintelligence: from Semiconductors to Superconductors