|
Your first live office hours session is Friday, August 14. Bring a real problem and we'll get it working together, or reply to this email with what you want covered and I'll prep it.
|
|
|
|
In AI this week
|
AI's everyday tier just got dirt cheap
Yes, it's Friday: this week's Thursday 3 aged an extra day, and the news made it worth it. Yesterday OpenAI cut the price of Luna, the budget model in its GPT-5.6 family, to a fifth of what it was, and took a fifth off Terra, the everyday mid-tier. OpenAI's own math on Luna: the quality that counted as frontier a year ago, at about six cents on the dollar, at nearly nine times the speed. And it isn't a chat-only toy; Luna can use tools and carry a multi-step job to the end. (OpenAI)
What it means for you: A cut this size doesn't just save money on what you already run; it changes what's worth automating at all. The high-volume work you'd never pay premium rates for (tagging every support ticket, summarizing every call, sorting every inbound request) just became cheap enough to hand over. If you priced an AI project this spring and the numbers said no, run the numbers again. They have moved twice since then, and only in one direction.
|
|
|
Google's new models are cheap on purpose
On July 21 Google released two new models, Gemini 3.6 Flash and Flash-Lite, and neither one is trying to be the smartest AI in the world. That's the point. Google built them for the repeatable work companies run at volume, its own words: the sweet spot of efficiency and quality for scaling AI workflows. The new Flash is faster and cheaper than the model it replaces and finishes the same jobs with less output. The two biggest AI companies spent the same ten days competing hardest at the cheap end of the market. (Google)
What it means for you: You now run a two-tier AI budget, whether you've noticed or not. The discipline is matching the model to the stakes: cheap tier for repeatable, checkable work (first drafts, summaries, sorting, research passes), premium tier for the calls where being wrong is expensive (the contract, the pricing decision, the board memo). That's one sentence of policy your team can adopt this week: default to the budget model, move up when the stakes justify it. The bill goes down and the quality of the work that matters doesn't.
|
|
|
Meta is handing out free AI agents that close sales
In June Meta opened Business Agent to businesses of all sizes: a free AI agent inside WhatsApp, Messenger, and Instagram that answers customer questions, recommends products from your catalog, books appointments, qualifies leads, and closes sales. More than a million businesses already run one. The part worth noticing is what happens behind the chat: businesses can connect the agent to hundreds of systems like Shopify and Zendesk, so it takes action on the business's behalf, inside rules and guardrails the business defines. (Meta) And on Wednesday's earnings call, Zuckerberg told investors this is the opening move of a much bigger enterprise push. (TechCrunch)
What it means for you: The chat window era of AI is ending; the era of agents doing scoped jobs under rules you set is here. You don't need to sell through Instagram for this to matter. The pattern is the thing: one narrow job, connected to the systems where the job happens, with guardrails on what the agent may do without a human. Pick one workflow a customer or employee waits on today (quote requests, appointment booking, order status) and ask what the first pass looks like with an agent handling it under rules you wrote. That's the shape most AI wins will take for the next year.
|
|
|
|
This week's workflow
|
|
A competitor-monitoring routine that pings you on real moves
|
|
|
WORKFLOW / THE +1
A competitor-monitoring routine that pings you on real moves
> OUTCOME
"Keep an eye on the competition" is the task that loses to every fire and every meeting. I set a recurring calendar block to review competitors weekly, and I declined my own meeting eleven weeks running. This workflow takes you out of the loop: a scheduled job watches your competitors' pricing and product pages, an AI layer summarizes what actually changed and whether it matters, and you get pinged only on a real move. Weeks like this one are the argument. Three platform shifts in ten days is more than anyone tracks by memory, and your market moves just as fast. You've effectively hired a junior analyst who never forgets to check, for about eight cents a run.
> SETUP
A page-change detector like Visualping on the pages that matter, Claude Code with a scheduled routine and web search for the scan layer, and a Slack channel or inbox for the digest. Plus your list of the two or three competitor pages where a change would genuinely alter what you do this quarter.
> WHAT YOU SHOULD DO NEXT
Point the watcher at your three highest-signal competitor pages and let it run. Then stop carrying the competition around in your head.
|
| |
↳ Full walkthrough _
|
|
|
Andrew
|
|
1:1 · OPERATOR TO OPERATOR
|
$1,000 · limited spots
|
Want to work through any of this directly?
I do a small number of Operator to Operator sessions each week. 60 minutes, screen-share, you bring the bottleneck and we leave with a concrete plan you can run that week. $1,000, pay then book.
|
|
|