AI Friday — Monday, August 31, 2026

AI at Work, With Better Guardrails

A practical guide to understanding ChatGPT Work

ChatGPT Work is powerful enough that its product boundaries can be hard to see. Simon Willison maps out what it is actually for and how its rapidly evolving features fit together—an excellent orientation before you hand it meaningful work. The HN discussion adds plenty of real-world questions.

Simon Willison

Check what Claude Code leaves in your commits

A reported default behavior appends a Claude session URL to commit messages and pull-request descriptions. That is worth checking in any shared repository: useful context can become an accidental sharing surface. See the original issue and the HN discussion before deciding how your team wants to handle it.

Hacker News

A simple prompt inventory for your AI footprint

What We Tell AI is a timely reminder that the most important AI data policy is often the one you make for yourself. Use it as a conversation starter for deciding what belongs in a prompt, what should be redacted, and what needs a human review step. The HN discussion is a useful companion read.

Hacker News

Google’s AI learning material is free

If you have colleagues asking where to begin, this free collection of Google AI skills is a low-friction on-ramp. It is more useful as a shared learning resource than another abstract “learn AI” recommendation.

There's An AI For That

Things to Build—and Lessons From Builders

Bolnee-Chat brings a self-hosted chatbot to a business site

This Show HN project offers a self-hosted route to adding a chatbot to a business website. It is a nice option to inspect if you want to prototype customer-facing AI without immediately committing to a third-party chat platform. Here is the HN conversation.

Hacker News

Artlist Flows turns creative AI work into a reusable canvas

Artlist Flows is aimed at the messy middle of creative production: turning a sequence of AI-assisted steps into a visual workflow you can revisit. That makes it interesting for teams producing repeatable video, image, or campaign assets.

There's An AI For That

What turns a vibe-coded app into a business?

On Behind the Craft, Amol Jain talks with Peter Yang about moving from an AI-assisted prototype to something customers will pay for. Expect the valuable non-code questions: choosing a problem, getting distribution, and not mistaking a fast demo for a durable product.

Behind the Craft

Caterpillar brings its automation experience to AI deployment

Caterpillar’s story is a useful business-side counterweight to the app-launch frenzy: years of putting autonomous equipment to work in demanding environments are informing how it approaches AI deployment. The takeaway for builders is that reliability and workflow fit still matter as much as the demo.

TechCrunch AI

Worth Reading and Listening To

No AI Fridays is a playful boundary-setting experiment

No AI Fridays proposes a deliberately analog day. Whether or not you adopt the rule literally, it is a worthwhile prompt for teams to notice which tasks AI truly improves and which ones benefit from uninterrupted human judgment. The HN discussion has the debate.

Hacker News

How I built this: a builder’s notes from Ben’s Bites

Ben’s Bites continues its “How I built this” series with a practical look at the work behind an AI product. Experience reports are often more useful than launch announcements because they expose the decisions, shortcuts, and tradeoffs that a polished demo hides.

Ben's Bites

A roundup of AI features actually worth trying

AI Daily Brief rounds up practical new features, from browser-oriented Claude capabilities to more flexible ChatGPT chats and Gmail connections. Good listening if you want a short “what should I test this week?” queue rather than another model leaderboard.

AI Daily Brief

The everyday habits behind better AI writing

Another useful AI Daily Brief episode, this one focused on five habits for making AI-assisted writing sound more like you and less like a generic draft. Especially relevant for anyone shipping client emails, proposals, or public copy with AI in the loop.

AI Daily Brief

Also

A plain-English look at diffusion language models — An approachable explainer for a model family that drafts and refines text differently from standard chat models.
How to build a diffusion language model — More technical follow-up for readers who want the implementation view.
Why agent security gets harder once agents can act — A useful companion to today’s commit-URL caution.
An AI agent accidentally deletes email — A concise case for narrow permissions and reversible actions.
Meta explores robots for data-center tasks — An interesting real-world robotics deployment story.
Lovable on apps that agents can use — A product-design conversation about building software for both people and AI agents.
Who let the agents in? — A quick read on the expanding set of agent-powered workflows.
How to start AI coding — Still a strong on-ramp for people looking for software-shaped problems to solve.
God’s Eye View — A visually striking public-signal intelligence demo.
The HN discussion on self-hosted business chat — Implementation questions and feedback on Bolnee-Chat.

AI at Work, With Better Guardrails

Things to Build—and Lessons From Builders

Worth Reading and Listening To

Also

Today’s Sources