Most people use AI for emails and not much else. The daily shows you what else it can do. As of this week, the AIgent membership is where you actually build it.
Here is what's inside. Build one is your own website, live, made with the AI you already have, with Mike walking you through every step on video. The other eight beginner builds land by November 15, and you keep every one. Live Builds is where Mike builds with members and you ask questions as you go. The Vault holds the full 100-prompt pack, the repos and the archive. And the members-only channels in the community open up.
$29 a month is the founding rate. After November 15 the price for new members may go up. You keep $29 for as long as you stay.
Seven days free, then $29 a month. One click to stop.
The daily stays free, every weekday, for everyone. Today's issue is below.
|
The AIgent Membership
Build your own website with the AI you already have Build one walks you through it step by step, with Mike on video at every checkpoint. It is a beginner build, and the site is yours when you finish. Seven days free, then $29 a month. One click to stop. |
The Drops
Repoclaude-plugins-official
38,885 stars · anthropics/claude-plugins-official
Anthropic's own curated directory of Claude Code plugins, split into internal (Anthropic-built) and external (vetted third-party) folders.
Here is the real problem it solves. Installing a Claude Code plugin used to mean trusting a random GitHub link with zero vetting. This gives you one install path with at least a submission bar behind it.
Quick start: run /plugin install {plugin-name}@claude-plugins-official or browse via /plugin > Discover.
Catch: Anthropic says it does not control what MCP servers, files, or other software a plugin includes, and cannot verify that they will work as intended or that they won't change. Read the plugin's homepage before you trust it.
AffiliateHubSpot
An official plugin directory is Anthropic drawing a line between what it vouches for and what it doesn't, and that same question, what actually shows up as trustworthy, is now playing out in how AI answers describe your own business. HubSpot AEO, a tool that shows how your business appears in ChatGPT, Gemini, and Perplexity answers, benchmarks you against competitors, and gives you recommendations to improve it. If your traffic is moving from search results to AI answers, this is where you measure it before you guess at it.
Use it for: seeing how your business shows up in ChatGPT, Gemini and Perplexity answers, and fixing the gap.
We may earn a commission.
MCPmailcheap
1,025 stars · marclou/mailcheap
A self-hostable newsletter app you pay for by emails sent, not by dormant subscriber count, with an MCP server built in.
Use it for: an agent-driven send pipeline, Claude, ChatGPT or Grok can write and review an issue through MCP, but sending itself is marked irreversible so the agent has to ask first.
Catch: it's brand new, pushed its first commit days ago. Early software, watch the issue tracker before you bet a production list on it.
|
The AIgent Membership
Every prompt in the pack, not one a day Start Here gives you one prompt a day. Members get all 100, tested on ChatGPT and Claude, plus the repos and the archive in The Vault. Seven days free, then $29 a month. One click to stop. |
Start Here
Turn your meeting notes into a list someone actually works from
One prompt today instead of a how-to. It is one of ours, tested on ChatGPT and Claude. Copy it, paste it into the free app you already use, and fill in three blanks: the messy notes you took, who was in the room, and today's date.
Here are my raw notes. They are unedited, out of order, and some lines are half sentences.
[RAW NOTES]
People in the meeting: [ATTENDEES]
Today's date: [TODAYS DATE]
Give me four sections, in this order.
1. Action items. One line each, in this shape: what needs doing, who is doing it, and the date it is due. Use the names I gave you. If my notes do not say who owns an item, write "owner not named" rather than picking someone. If my notes give no date, suggest one and mark it "suggested".
2. Decisions made. One line each, saying what was settled. Only include things my notes show were actually decided, not things that were discussed.
3. Open questions. Anything left hanging, written as a question, with the name of the person most likely to be able to answer it, or "owner not named" if the notes give no clue.
4. Leftovers. Anything I wrote down that does not fit the three sections above. Quote my own words back to me rather than summarising them.
Do not invent an action item that is not in my notes. Empty sections are fine when the notes have nothing for them. At the end, if the notes give you grounds for it, tell me which single item is most likely to slip and why, and mark that as your guess.
The line that makes it work: "If my notes do not say who owns an item, write 'owner not named' rather than picking someone." Most AI summaries guess an owner to look complete, and a guessed owner is worse than no owner, because someone will quietly assume the wrong person is handling it. Forcing the honest gap into the output is what makes the list trustworthy enough to actually send to your team.
TRY THIS: open your last meeting's notes app right now, paste the raw mess into the prompt exactly as you wrote it, half-sentences included, and see what comes back.
This is one of 100 in the AIgent prompt pack. The other 99, plus guided builds with Mike on video, are inside the AIgent membership: Join the AIgent membership. Seven days free, then $29 a month. One click to stop.
|
Recommended
Once the real action items are sorted from the noise, the next fight is keeping them from scattering back across seven different apps and three chat threads. ClickUp, one workspace for tasks, docs, and chat brings your tasks, docs, chat, and goals into one place so a small team stops switching between seven apps, free to start. We may earn a commission. |
Frontier Signals
Meta and Apple, usually rivals, worked together on the launch of Muse Spark Open. Meta says it is the first model in its class to run on a desktop, using two Apple M5 Ultra Mac Studios with the MLX framework. For operators, it means a serious open-weight model with real Apple Silicon optimization behind it. (The Verge)
Open-weight models now carry 71.5% of AI Gateway token traffic, but only 23% of the spend. Vercel's October 2026 AI Gateway production index, covering September, shows customers increasing budgets on both sides of the market, and frontier spend has grown tenfold since December. (Vercel)
Anthropic is warning that small models could soon copy themselves across internet-connected devices. It says they could spread across devices such as smart TVs and cameras, creating botnets that are hard to shut down, though it has not shown this through commercial data centers. If you're deploying agents on anything with network access, this is the threat model to start designing against now. (CBS News)
A hack of store refrigeration controllers hit roughly 700 supermarkets, with French authorities involved and other European countries also reportedly affected. Attackers remotely took over internet-connected controllers, pushed fridges as high as 50°C, spoiled food, and in some cases set off electrical fires. The lesson for anyone wiring agents into physical operations: the weak link is often an outdated, internet-exposed controller. (Ars Technica)
Booz Allen tested AI models in an isolated OT lab, and in one test they got from a perimeter compromise into an industrial control network in just over 16 minutes. The report says the models navigated unfamiliar industrial systems, adapted when blocked, and moved physical equipment within minutes. If your physical infrastructure touches the internet anywhere, that's now a live attack surface. (SecurityWeek)
|
Recommended reading
If you like The AIgent, a small group of operator-tier publications worth your inbox: see the shortlist. |
What are you building with right now?
Before You Go
Hit reply and tell me which Drop you're installing first, the plugin directory or MailCheap.
Anthropic built the one directory it vouches for and left the rest of the ecosystem in place. That's usually how platforms mature: a better default alongside what already exists.
See you Monday.
Before you go: we started a room for people actually building with agents. Post what you shipped this week. Join the community →
|
Want to reach builders shipping with AI every weekday? Advertise with The AIgent. The AIgent stays free for everyone, and reader support keeps it that way: Support The AIgent. |
