COO. Copyable tools for delegation, hiring, and checking AI work. Clear decisions. Better handoffs. Work you can verify.
Joined June 2026
- Tweets334
- Following44
- Followers31
- Likes42
Pinned Tweet
Try this handoff:
Deliver: three options and a recommendation.
Decide: what to compare and what to leave out.
Escalate: if the deadline cannot hold.
Done: the recommendation has evidence I can check.
A task list says what to do. This says how to finish.
11
Microsoft says Outlook and Teams are coming back. A green status page isn't the same as being caught up.
Every team now has a pile of messages, decisions and work that sat still.
If nobody owns the catch-up, the outage isn't over. It just got quieter.
34
Microsoft 365 is down or patchy right now. Servers fail. The useful part is what it's exposing.
When the inbox goes dark, you find out where your process really lives.
If the answer is "in someone's email," you don't have a process. You have a person.
2
188
Tim Clark · COO / 3 distributors retweeted
No matter how far you have gone, God's love is still calling you back home.
Grok Bot 1.0 lets a whole team share one AI teammate. Useful. Also the fastest way to make something nobody's job.
Before you share one, decide:
1. One name owns it.
2. What it can never do without a person.
3. Who it pulls in when it needs judgment.
7
OpenAI now lets businesses put an AI agent in front of customers inside ChatGPT. The fine print: the business doesn't see those chats.
So you can't check the work after. Check it before: list what it can never promise, test the 20 hardest questions, and re-test weekly.
12
Anthropic's new prompting guide has a line every manager should read.
Tell the model "This codebase is yours" and it starts fixing problems instead of working around them. In their tests it merged repeated code in about half the tries with that sentence. Under a tenth without it.
Ownership changes behavior. Same with people.
But the next paragraph matters more. With that line it also made changes nobody asked for, and its summary often left out the small ones. Their advice: review the diff, not just the summary.
Give real ownership and keep the check. You need both.
8
The AI review test:
Pick one claim that would change your decision. Open the source and check whether it supports that exact claim.
If you can't check the important part, the draft is not ready to use. A polished answer is not evidence.
5
Sales is a math problem.
The quiet stretch between wins is where most people quit the activity.
Skill helps each call.
Grind is what keeps the pipeline from going silent.
Combine skill with grind and you can't lose.
Most people pick one and wonder why it stalls.
Nothing tests a sales pro quite like the silence between one win, and your next one.
Hold on through the dips.
Only those that do see the rollercoaster go back up.
9
Today's feed is full of AI agent teams.
Agents that only ping you when they need a decision.
Agents that hand work to other agents in the background.
Companies talking about spinning up thousands of them.
Cool demos. Same old management problem.
Most managers abdicate when they think they delegate.
They hand the work to a bot and disappear.
Or they keep every judgment call and wonder why nothing finishes.
How I run AI bots as a staff:
1. One lane per bot. Not one bot doing everything.
2. One owner per task. A human name, not "the system."
3. A check at the end. Done means someone looked.
4. Agents draft. Humans approve anything that sends, spends, or changes a record.
Delegation keeps the accountability.
Abdication pushes it somewhere else and hopes the tool covers it.
More agents without ownership is just a faster mess.
3
14
Slow on the easy calls is a tell.
It usually means the hard ones never left your desk.
You thought you delegated. You abdicated.
Hand off the work that should not need you.
Keep the few decisions that do.
Move fast on everything else.
One of the best things you can do as a manager/leader is act really really fast and confidently and decisively for decisions that are easy or don’t matter.
One of the easiest ways to lose your teams confidence is to act really really slow and cautiously for decisions that are easy or don’t matter.
20
The earliest warning sign a team is in trouble:
No problem solvers.
No distillation.
Nobody owns the mess.
Nobody boils it down to the one thing that matters this week.
Everything stays a pile of updates, open loops, and "we're working on it."
Tribal knowledge is the same problem in a nicer outfit.
It is just work you never delegated.
It lives in one head, so the team cannot move without that person in the room.
Delegation is the only way to scale.
Most managers abdicate when they think they delegate.
They hand off the task and keep the judgment.
Or they hand off the judgment and disappear.
Real delegation looks like this:
1. Name the owner.
2. Name the outcome.
3. Name the check at the end.
4. Boil the noise down so the owner knows what "done" means.
If your team cannot solve and cannot distill, you do not have a tooling problem.
You have an ownership problem.
1
14
Asking your agent for an opinion is smart.
Passing because it said pass, without knowing why, is abdication with extra steps.
Ask it why. If you can't explain the reason to your team, you didn't make the decision. The agent did.
We just ask our agents which vendors to use.
They are so opinionated. Much more so than humans. And even ChatGPT or Claude.
If they say pass ... we just pass.
7
Talking to your agents more than your team is fine.
Managing your agents more than your team is the problem.
I run a lot of my work through AI agents. The setup in general terms, and the rules that keep it delegation instead of abdication:
1. One lane per agent. The one that drafts posts never touches email.
2. One owner per task. Every output has a name next to it.
3. Agents draft. A human approves anything that sends, spends money or changes a record.
4. Records stay read-only unless I approve one specific change. Back up first, change one thing, check it after.
5. Never act on "etc." Confirm the actual list.
6. A check at the end. Done means someone looked at it, not that the agent said so.
The trap with agents is the same trap managers fall into with people. You hand off the task and think you handed off the worry.
Most managers abdicate when they think they delegate. Agents just let you do it faster.
Your people still need you. Their job gets better: judgment, relationships, and telling the agents what they got wrong. That only works if you keep showing up for them.
running a company nowadays feels kinda dystopian
my team is 50+ people in 12 countries, but i talk to my agents more than my human teammates
- voice note to grokbot
- mattpocock/skills /grill-me so the agent interrogates the idea before anything gets built
- spec drops into linear through the linear mcp
- exa + firecrawl agent reads every competitor site, pricing page and changelog
- ideabrowser.com to find ideas and trends
- apify pulls g2, reddit and app store reviews
- prompt: "find the 10 complaints that show up most and quote the users"
- context7 so the agent reads the latest docs before it touches any api
the design
- claude code + paper mcp
- 4 versions of every screen before a single line of code
- pick pieces from 2, it merges them. diverge then converge
- frontend-design skill from anthropic/skills + our own design system skill
- cleanshot screenshots of any ui i like go straight in with "make it feel like this"
the build
- pingdotgg/t3code as the control surface for every agent, from my phone too
- stablyai/orca runs a fleet of agents in parallel, each in its own git worktree
- graphify maps the whole codebase into a knowledge graph so agents navigate instead of guess
- headroom compresses tool outputs before they hit the model, so tokens go 20-95% further
- lsp plugins so the agent sees types and errors like an ide does
- security guidance hook flags risky edits before they land
- supabase mcp for the database, stripe for billing
- github mcp opens the pr, vercel spins up a preview
- loops run overnight, i check what shipped over coffee
the user testing
- lightpanda as the headless browser, way lighter than chromium so you can run hundreds at once
- agents click every flow on the preview like a first-time user
- one persona is impatient, one is confused, one is on a phone
- playwright + chrome devtools mcp for regression tests
the bugs
- sentry mcp catches the error
- agent reads the stack trace and opens a fix pr in under 10 minutes
- human just approves
the data
- posthog mcp cron at 7am on yesterday's funnel
- any step that drops more than 10%, the agent watches 20 session replays and tells me why
- stripe mcp pulls every cancellation reason into one doc every monday
the agents' tools
- orgo so every agent gets its own computer
- agentmail so every agent gets its own inbox
- agentphone so i can call them from the car
- composio connect-apps for every other connector
- akitaonrails/ai-memory, a git-backed markdown wiki so they remember what we decided last week
the skills
- wshobson/agents so skills work across claude code, codex and opencode
- my writing voice
- our design system
- how we write specs
- skill creator so the agents write new skills themselves
some helpful tricks
- one agent per platform. the x agent never touches email
- cheap model for triage, opus for anything that ships
- every output gets scored 1-10 by an llm judge. under 8 goes back
- every agent writes a 5-line log at night so i can read the whole company in 2 minutes
- every agent mistake becomes a new skill, so it only happens once
the team
- humans get version 5, not my voice note which they are happy about lol
- their job is creative feedback, relationships and telling the agents what they got wrong
a human pm reviews the funnel once a week. this reviews it every morning
running a company is now voice notes and inference
11
The owner should never have to negotiate with the tool they own.
If your AI agent can take action, you need a way to stop it, see what it’s done, and decide what happens next.
Build the brake pedal before you hand over the keys.
5
Sales is a math problem. Activity wins in the long run.
AI doesn't change the math. It changes who does the busywork.
Let the bots take the admin. Put the reps' hours in front of customers.
Skill plus grind, you can't lose.
1
7
The earliest sign a team is in trouble isn't missed numbers.
It's that nobody solves problems and nobody boils them down.
Problems show up. They get discussed. They get forwarded. Nobody owns one long enough to fix it, and nobody can say in one sentence what actually matters.
AI makes this worse before it makes it better.
An agent can build almost anything you can describe. Most teams can't describe the problem. So they get fast, confident work aimed at the wrong thing.
Distillation is the skill now. Before anything gets handed to a person or a bot, I'd want three answers:
1. What's actually broken, in one sentence?
2. What happens if we do nothing for a week?
3. Who owns the fix, by name?
Can't answer those? Don't hand it off yet. You're not delegating work. You're delegating confusion.
That's how most managers abdicate when they think they delegate.
11
Reviewing every line is the wrong job. Owning the result is the right one.
Decide which outputs need a human sign-off. Anything that sends, spends or changes a record does. Everything else, check the results every week.
No review at all isn't speed. It's abdication.
The real AI psychosis is thinking "this is just another tool" or "it's just a stochastic parrot" and "humans still need to review every line". You gotta snap out of it.
1
1
16