Data isn't the new king. It has always been king.
Every AI platform is racing for the same thing: the transaction data sitting in your SAP system. Models are getting cheaper. Your business context isn't.
linkedin.com/pulse/data-isnt…
12
One Bot, One Job
How I cleaned up the AI assistants I spun up too fast, and the one rule none of them gets to break.
Over the past few weeks I set up a lot of AI assistants: email, X, LinkedIn outreach, travel between my two home bases, and a few for the software side.
Each one made sense the day I created it. Together, they were a mess.
Today I approved a fix. It's simple, and it's something 30-plus years in business taught me long before AI showed up: every person needs one clear job. That goes for bots too.
THE PROBLEM: I HIRED FAST AND NEVER WROTE JOB DESCRIPTIONS
I asked one assistant to review the others and report what each was actually doing. Here's what it found:
• Email was split between two bots. Both were working in my inboxes. Neither had a clear lane.
• Technical work on my internal app was spread across four bots.
• A meeting requirement got drafted twice. Two bots wrote it up from the same call, on the same day.
• Outbound sales sat in two places. One bot owned LinkedIn outreach. Another wrote a go-to-market plan for outreach on its own.
• One bot had sat idle for about a month. It got stuck waiting on a sign-in and nobody noticed.
• There was a blank bot nobody had ever used.
None of this was the bots' fault. I'd done what a lot of founders do with new hires: brought them on, said "help out," and walked away.
THE FIX: ONE BOT, ONE JOB
The new roster has 9 working bots, each with a single job:
• Courier: correspondence and research. Drafts replies to real people in email and Slack, and handles one-off research and writing.
• Inboxer: mailbox hygiene. Files junk, rescues real mail, watches for spoofing, and flags what needs me. It doesn't write replies.
• Dex: engineering lead. Owns the code for my internal app, an AI-native task and CRM system, plus its pull requests. Also owns the daily decision packet (more on that below).
• Scout: AI and tech scout. Watches for new models and tools and writes short "should we adopt this" memos. Anything worth building goes to Dex.
• Minutes: meeting notes. Turns calls into summaries, action items, and draft requirements. It's now the only bot that writes requirements.
• Pipeline: pipeline outbound. Account research, warm intros, and outreach drafts in my voice.
• Postbo: X and personal brand. Posts, threads, and images.
• Wayfinder: travel desk. Flights, trip options, and calendar holds for my back-and-forth between two home bases.
• Gridiron: sports and fantasy. Personal. Weekly picks and lineup reminders. Not everything has to be work.
The other bots now hand their findings to the owner instead of doing the work themselves. If Scout finds a better search tool, it writes a memo for Dex. It doesn't go off and start building.
THE FLOW: RESEARCH, THEN DRAFT, THEN DECIDE
Work moves in one direction.
Research comes from the bots closest to the information: Scout for tools, Pipeline for accounts, Courier for one-off questions.
Drafting belongs to whichever bot owns the channel. Courier drafts email and Slack. Postbo drafts X posts. Pipeline drafts LinkedIn messages. Minutes drafts meeting notes and requirements. One writer per channel means no more duplicate drafts.
Deciding is my job, and it happens in one place.
By 6:45 AM, every bot drops a few lines into a shared folder: what it needs me to decide, plus anything I should know. Dex merges all of it into one 7 AM decision packet.
Decisions come first. Each one is written as an exact action I can approve: "Send this reply." "Post this thread." "Book this flight." Not "you might want to look at the travel situation."
An end-of-day summary works the same way.
THE HARD RULE
Nothing leaves the box without my explicit OK for that exact action.
That means no sending, posting, publishing, booking, buying, or submitting until I've approved that specific thing. A general "go ahead with outreach" doesn't count. And nothing gets deleted.
This is the part I'd hold onto even if everything else changes. Bots are fast, and fast mistakes in your inbox or on your public accounts are expensive.
WHAT'S STILL ROUGH
I approved this today, so I'm not going to tell you it works yet.
Here's where it actually stands:
• The handoffs went out this afternoon. Each bot got its job description and is updating its own profile and routines to match.
• The first 7 AM packet is tomorrow. That's the real test: does one merged list beat nine bots pinging me all day?
• 7 AM follows me. It's 7 AM local time wherever I happen to be. When you split your life between two places, that has to be decided up front.
• The idle outbound bot needs me. It's waiting on one sign-in from me before it can restart, and I still have to decide whether I want LinkedIn outreach at all.
• The blank bot is still there. I haven't decided whether to retire it.
• The inbox cleanup might be too noisy. It runs 6 times a day. Most runs flag nothing for me, so cutting to 3 is on the table.
• No saved skills yet. Five repeatable workflows were identified. Two are drafted. None are saved.
I'll report back once it's actually been running for a while.
WHAT I'D TELL OTHER FOUNDERS
Treat bots like hires. If you wouldn't bring on a person without a role, don't do it with a bot.
Write one job description per bot. One job, the sources it owns, what it delivers, and what it needs your approval for. If you can't write it in a line, the job is too big.
Decide who owns each source. One bot per inbox, repo, or social account. Shared ownership is how you get two bots in the same mailbox.
Run one morning decision queue. Don't let every bot ping you all day. Make them line up, and make every item an exact action you can approve or reject.
Turn repeated wins into saved skills. When a bot does something well, save the method so it isn't reinvented every time.
YOUR TURN
If you were hiring your first AI assistant tomorrow, what's the one job you'd give it? And what would you never let it do without asking you first?
Tell me in the replies.
(Header image is AI-generated.)
🤖 Made with AI
4
398
🤖 Made with AI
From one man called out of Ur, God built a nation. Abraham's family tree produced the twelve tribes of Israel, and from the tribe of Judah came the line that led to Christ. Ishmael, Esau, Moab, and Ammon also appear here. The genealogy of Abraham shaped the entire ancient Near East.
13
Underwater humanoids tired of leaving beneath Easter Island decide to show off their lighter side.
Watch until the end
🤖 Made with AI
30
What's one thing in this image that shouldn't be there?
The superintelligence war won't be won by the best demo.
It'll be won by whoever ships working AI into factories, grids, and defense first.
The U.S. can win that. Only if we stop confusing announcements with deployment.
AI concept image, not a real event.
🤖 Made with AI
20
If SMX built a Legends League tomorrow, who wins it?
One name. Settle it in the comments.
AI concept video, not a real race.
#MXLegendsLeague #Supercross #Motocross #WhoStillHasSpeed
🤖 Made with AI
2
25
New hit: Don’t Let the Lies divide us
suno.com/s/AfsYVdgV4Cfosgv1
1
20
Did you know some energy drinks add L-theanine, the same compound that’s naturally in tea? It apparently takes a bit of the jittery edge off you get from a cup of coffee.
Cool and refreshing pick-me-up you can buy in bulk at Costco.
🤖 Made with AI
2
26
Ok I’m going to tell the truth, this is incredible, but I think it is worth the slightly more or less to just go somewhere similar. Bali, Kaua’i, etc.
Ever wonder what $1874.40 of Opus 5.5 tokens looks like?
Wonder no more.
I highly recommend watching the whole video on 1080p on a large screen with sound on.
Pay attention to all the little details.
- Birds flying and diving into the ocean.
- Cloth, signs and lights swaying in the wind.
- Crabs scuttling along the shore and burrowing when you get close.
- Fish swimming alongside the whale.
- Lights illuminating the dock at night
Then zoom out and see the entire island. Never dipping below 60 FPS at 1440p resolution.
Yes, there's a few bugs and visual artifacts, some textures need improving, the shorelines waves sometimes look funny, but these are so trivial to fix at this point.
FYI I'm on the $200 subscription plan. This used 59% of my weekly usage and took 2h 7h of API time using multiple subagents, about 8h in real-time.
There was extremely little technical direction here. 99% of my prompts were "Add X and Y" or "This looks weird, make it better".
It's a great time for hobbyists, bad time for professionals. This experiment has further cemented my view that technical creatives are about to experience a massive disruption.
28
Well maybe a little of both. I definitely feel way more productive, but burnout is a real potential because I feel like in order to keep pace with the AI changes you have to think in a 24/7 pattern.
That can cause real burnout if you are not careful. I recommend do not change your walk the dog and gym routine.
The work will be there when you get back.
AI is suppose to make use faster, but the bar for "perfect output" just increased.
Does AI make you feel:
0%More productive than ever
0%Quietly burning out
0 votes • Final results 1
81
David Caspillo retweeted
AI is suppose to make use faster, but the bar for "perfect output" just increased.
Does AI make you feel:
0%More productive than ever
0%Quietly burning out
0 votes • Final resultsLots of opportunity here to not only secure the border but to increase state by state revenue. Good thread!
The US-Mexico border and the Canadian-US border are the pulse of our economy, but antiquated systems with 150-200% error rates are clogging the veins.
A mere 10-minute reduction in wait times can trigger a $2.2 BILLION annual trade boost. That is the power of the BWT Project.
We are revolutionizing border management with a high-accuracy, infrastructure-less AI-hybrid system (89-93% predictive accuracy). Our lean approach delivers enterprise-grade results at a fraction of the cost of heavy surveillance hardware.
The "Funding-to-Impact" Alpha:
✅Seed Funding Ask: $15M
✅Year 5 Revenue Projection: $200M+
✅Target: 167 US land Ports of Entry (POEs)
Strategic Moat: Seamless Google Cloud & Maps integration for rapid, global scale.
By solving for efficiency rather than just surveillance, we’re not just building a company we’re putting billions in economic potential for North America. 🇺🇸🇲🇽🇨🇦
#BWTProject #AI #BorderTech #EconomicGrowth #Innovation #GoogleCloud
1
54
I don’t see it
Hey @grok put the same clothes, hat and beard from the man on the right side to the guy on the left side
36
I did that once and the other person standing waiting to select a seat said that was the new CIO you are presenting to. 😰
This quoted post is unavailable.
35


















