Daily AI Zine · No. 036

Today's Big Thing

The one thing that matters

Use it

ChatGPT Work and Codex turn launch pressure into the practical lead

The day’s most useful signal is execution, not a new model.

OpenAI published a customer case showing Stampli using Codex and ChatGPT Work to compress weeks of launch production into days while facing a fixed deadline and limited design resources. That is not a broad platform announcement, so today’s lead is a useful shift rather than a major development. The pattern matters because it frames AI work as deadline rescue: gather product intent, draft launch assets, implement the supporting pieces, and keep humans on approval. Verdict: use it for bounded launch work with a named deadline, not for open-ended brand strategy.

T1OpenAI News08/20

My AI Ecosystem

Your actual stack

Use it

GitHub Code Quality gets cleaner Actions accounting

GitHub says the dedicated workflow path for GitHub Code Quality CodeQL actions is now generally available. Workflow history and Actions usage reports can now separate code quality runs from other automation. For any repo using GitHub Actions, this makes quality scanning easier to budget, inspect, and explain during delivery.

T1GitHub Changelog08/20

Use it

CodeQL sharpens Actions and JavaScript scanning

CodeQL 2.26.3 adds JavaScript, TypeScript, and Vue source modeling and improves the accuracy of several GitHub Actions queries. CodeQL is GitHub’s static analysis engine, meaning it scans code for risky patterns before release. This is practical repo hygiene for web projects and automation-heavy workflows.

T1GitHub Changelog08/19

Watch it

Hot signal: Agent Lightning trains agents inside the harness

Agent Lightning v1.0 is a research signal, not a tool to deploy today. The useful idea is that the agent harness, meaning the system that manages tools, context, and control flow, is becoming part of training rather than just a wrapper around the model. Watch this because coding agents increasingly fail or succeed based on their operating setup.

T1Hugging Face Daily Papers08/17

CoWork Corner

Claude CoWork, day to day

CoWork: separate thinking from execution

Claude CoWork is the workspace Adrian uses for persistent projects, operational records, and AI-assisted workflows. No CoWork-specific product change showed up today, but the useful workflow is still grounded in Claude’s MCP direction: keep the durable plan, decisions, and handoff notes in CoWork, then send bounded tool work through an MCP-capable execution lane. MCP means Model Context Protocol, a standard way for AI tools to connect to external systems. The rule is simple: CoWork holds intent and memory; execution tools return evidence and summaries.

T1Anthropic Engineering Blog

GPT Desk

OpenAI, ChatGPT, Codex

Use it

Use ChatGPT Work and Codex as a launch-production pair

What changed: OpenAI’s Stampli case shows ChatGPT Work and Codex compressing launch production from weeks into days when deadline pressure and design constraints collided. Why Adrian should care: Goodsense, SET, and Grey Group often need proposals, pages, decks, and production documents faster than a full design cycle allows. What to do: pick one live or recent offer and run a two-lane test, ChatGPT Work for messaging and asset structure, Codex for the delivery page or repo changes.

T1OpenAI News08/20

Test it

Turn OpenAI’s privacy work into a routing test

What changed: ITmedia reports OpenAI plans Private Safety Processing while maintaining Zero Data Retention, and OpenAI has separately reaffirmed ZDR for eligible API customers while previewing the safety approach. Why Adrian should care: Grey Group OS needs a clear split between client-sensitive API work and lower-risk automation. What to do: write one routing rule today that says which Goodsense, SET, or Street Attack Japan tasks may use OpenAI APIs, which must stay out, and who approves exceptions.

T2ITmedia AI+08/20 T1OpenAI News08/19

Small Money Systems

Small, repeatable, real

Use it

Small system: code-quality checkup for small sites

System: create a lightweight GitHub Code Quality checkup for small web projects. Customer: local businesses, creators, production microsites, or campaign teams already using GitHub and Netlify. Offer: they receive a short risk report, top fixes, and a plain-English release-readiness note. Price: ¥30,000 for a first checkup. Existing assets: Grey Group OS, GitHub delivery habits, Netlify deployment experience, and Claude Code or Codex review workflows. AI workflow: GitHub CodeQL and Actions produce the scan trail, then Codex or Claude Code summarizes issues and drafts fixes. First action: run the workflow on one non-sensitive repo. Repeatability: convert the report into a monthly maintenance retainer. Effort: half day. Expected value: fewer release surprises and a repeatable maintenance offer.

T1GitHub Changelog08/20 T1GitHub Changelog08/19

Build Next

Deployable now

Use it

Build a separate code-quality release lane

What to build: a repo workflow that runs GitHub Code Quality separately from deploy checks and posts a short release-readiness note. Why now: GitHub says Code Quality now has its own Actions workflow path, and CodeQL 2.26.3 improves Actions queries plus JavaScript, TypeScript, and Vue modeling. Effort: half day. Expected impact: cleaner release decisions and fewer overlooked automation or web-code risks. Dependencies: an existing GitHub repo, GitHub Actions enabled, and at least one web project where CodeQL scanning is appropriate.

T1GitHub Changelog08/20 T1GitHub Changelog08/19

Try This Today

One action, right now

Pick one active GitHub repo and separate the code-quality workflow from deploy automation. Let the scan produce a short release note: what changed, what risk appeared, and what needs a human decision before the next agent touches the repo.