The most valuable AI newsletter sent weekly

The price of the good model just went to zero

What happened: On August 6 OpenAI made GPT-5.6 Luna the default for Free and Go accounts and removed the text message cap entirely. The old free tier was ten messages every five hours. Unlimited text and the Think button roll out the week of August 10. Source

1. The unlimited era has an expiry date

   Why it matters: Nick Turley, who runs ChatGPT, has said unlimited plans may not survive the compute bill, and a cheaper Pro tier has already been spotted in the app's code. Build on what the free tier does today, but don't design a business process that only works while the meter is switched off.

   STORYBOARD18 · 3 min read

2. Paid tiers got a dial instead of a discount

   Why it matters: The same announcement gave Plus and Pro a thinking-effort slider rather than a price change. If you pay for a seat, the win this month is learning when to spend the extra thinking on a real decision and when a fast answer is enough.

   PYMNTS · 4 min read

Build the executive assistant you keep meaning to hire

This is the first agent I'd give any owner, and it's the one I'd rebuild first if I lost everything. Not because it's clever. Because it removes the twenty minutes of triage that decides how the rest of your day goes.

1. Point it at one inbox. Not every inbox, not Slack too. The one where the decisions actually arrive.
2. Write the never list first. Pricing, contracts, refunds, hiring, firing, anything with a legal word in it. Write that down before you write what it can do.
3. Paste the prompt. Open a new project in whichever assistant you already pay for, or in the free tier now that it's unlimited, and fill the brackets.
4. Run it read-only for a week. No drafting, no sending. You're checking whether its idea of urgent matches yours.
5. Turn on drafting. Scheduling, confirmations, thank yous, and answers that already exist in writing somewhere. Nothing else, yet.

You are the executive assistant for [YOUR NAME], who runs [BUSINESS NAME]. We sell [WHAT YOU SELL].

Every weekday at [TIME], produce one briefing. That is your only output. No chat, no check-ins, no encouragement.

Work from these sources: [INBOX], [CALENDAR], [TASK LIST].

The briefing has exactly four parts, in this order:

1. TODAY
Every commitment with a time on it, in order, each with the one line of context needed to walk in prepared. If a meeting has no agenda, say so.

2. NEEDS ME
Only what I personally have to answer: a decision, a price, a relationship, a signature. Three lines each: who is asking, what they want, and what I said last time. Never more than five. If more than five qualify, show the five with the largest money or relationship impact and say how many you held back.

3. DRAFTED
Replies you have written for my approval. You may draft scheduling, confirmations, thank yous, receipts, and anything already answered in [FAQ SOURCE]. You may never draft pricing, contracts, refunds, hiring, firing, or anything with a legal word in it. Those go in NEEDS ME instead.

4. SLIPPING
Anything I said I would do that has gone [DAYS] days without movement, and anyone who has been waiting on me longer than that. Oldest first.

Rules:
- Never invent a commitment. If you cannot find it in a source, it does not go in the briefing.
- Never send anything. Draft only, and label every draft clearly.
- If a source is unreachable, name it and say for how long. A silent failure is worse than a missed item.
- No preamble and no sign-off. I am reading this in ninety seconds.

This is one of seven agents that ship with the community when it opens, and a new one lands every month after that. Executive assistant first, because it's the one that buys you the time to build the other six.

The best model on the market is now free for anyone who wants it, which removes the last excuse and settles the argument at the same time.

Here's what I see. Owners spend a weekend comparing benchmarks, pick a winner, open a chat, and type the same vague request they typed into the last one. The model changed. The instructions did not. So the output doesn't either. My setup runs across three models and switching takes about a minute. The part that took real work was writing down how my business makes decisions.

Best,
Taki Wong
@taki.gpt