Rough draft first.
Waiting until you can do it well is a polite way of not starting.
You can't edit a blank page. The quality shows up in pass two and three.
What would you start today if it was allowed to be bad?
Waiting until you can do it well is a polite way of not starting.
You can't edit a blank page. The quality shows up in pass two and three.
What would you start today if it was allowed to be bad?
0
0
0
4
Keep the receipts.
Progress is hard to see from the inside. Every day looks like the last one.
Write down what got easier. On a bad day, that list is proof.
Where would your receipts say you've grown this year?
Progress is hard to see from the inside. Every day looks like the last one.
Write down what got easier. On a bad day, that list is proof.
Where would your receipts say you've grown this year?
0
0
0
9
Plant it anyway.
Effort shows up late. You do the work now, and for a while the ground looks the same.
That's when most people stop. Not because it failed, but because nothing came up yet.
Roots grow first.
What are you planting that hasn't come up yet?
Effort shows up late. You do the work now, and for a while the ground looks the same.
That's when most people stop. Not because it failed, but because nothing came up yet.
Roots grow first.
What are you planting that hasn't come up yet?
0
0
0
9
Say no to protect the yes.
Every yes costs time you meant for the thing that matters.
Most of what crowds it out isn't bad. A quick call. A small favor. Each one is reasonable. Together they eat the week.
What do you need to say no to this week?
Every yes costs time you meant for the thing that matters.
Most of what crowds it out isn't bad. A quick call. A small favor. Each one is reasonable. Together they eat the week.
What do you need to say no to this week?
0
0
0
10
Make the right thing easy.
Willpower runs out. Setup doesn't.
Shoes by the door. Phone in another room. Notebook already open.
The most disciplined people aren't resisting more. They've set things up so they resist less.
What could you make easier to start this week?
Willpower runs out. Setup doesn't.
Shoes by the door. Phone in another room. Notebook already open.
The most disciplined people aren't resisting more. They've set things up so they resist less.
What could you make easier to start this week?
0
0
0
10
Gemini 4 Carbon is #1 on Artificial Analysis' new index. The bigger news: 60% of that index now runs on private tests, because public ones leak into training.
Do the same for your agents. Keep a private set of real tasks, graded on "would we merge this."
What's in yours?
Do the same for your agents. Keep a private set of real tasks, graded on "would we merge this."
What's in yours?
0
0
0
13
Spawn's founder woke up to 700 questions from his agents. Most he'd already answered.
They built judges on two years of recorded decisions: cite the record, send it back, or hold for a human. He stepped in 27 times in 11,000+ verdicts.
How much of your judgment is written down?
They built judges on two years of recorded decisions: cite the record, send it back, or hold for a human. He stepped in 27 times in 11,000+ verdicts.
How much of your judgment is written down?
0
0
0
14
Practice the weak spot.
Drilling what you're already good at feels productive. Mostly it's rehearsal.
Ericsson's research on experts found the gains came from deliberate work on the specific things they couldn't do well.
What's the weak spot you keep working around?
Drilling what you're already good at feels productive. Mostly it's rehearsal.
Ericsson's research on experts found the gains came from deliberate work on the specific things they couldn't do well.
What's the weak spot you keep working around?
0
0
0
14
Theo: "If I have to go to a dashboard to use your product, I'm not going to use your product."
Same day, Apple put MCP in Keynote, Pages, and Numbers.
A dashboard is a human-only API. The next user of your product is an agent. Ship the API first and treat the UI as a view.
Same day, Apple put MCP in Keynote, Pages, and Numbers.
A dashboard is a human-only API. The next user of your product is an agent. Ship the API first and treat the UI as a view.
0
0
0
15
Last time, @rUv razzled and dazzled us with his seemingly impossible feats.
This time... we build. We share. We collaborate.
Bring your ideas, your questions, and whatever you're working on. Let's put our heads together and do great things!
Join us for the next @AgenticsOrg cago meetup:
November 12 | 6–8 PM Central
@WorkboxCompany Chicago – River North
Register: luma.com/a3egmqdb
#AgenticsFoundation #ChicagoTech #AI
This time... we build. We share. We collaborate.
Bring your ideas, your questions, and whatever you're working on. Let's put our heads together and do great things!
Join us for the next @AgenticsOrg cago meetup:
November 12 | 6–8 PM Central
@WorkboxCompany Chicago – River North
Register: luma.com/a3egmqdb
#AgenticsFoundation #ChicagoTech #AI
0
0
0
74
The Grok Bots I actually run my two businesses with.
Tote Sweet Home rents reusable moving totes in the Chicago western suburbs. Apex Insights helps independent restaurants. This is the fleet behind both, and where each bot has to stop and ask me.
1. Watson, the goals hub
I give a goal to one bot. It routes Tote work to the Tote GM and Apex work to the Apex GM, and they report status back up.
Payoff: one place to ask "where are we?"
2. The commitments chase
Weekdays around 4 PM it reads my Plaud meeting notes, logs what I said I'd do, and drafts head starts on the oldest items. It also tracks what other people said they'd get me.
Payoff: a promise made in a meeting doesn't quietly die.
3. Morning day prep
Weekdays around 7 AM it turns my calendar and open commitments into top priorities, conflicts, and open blocks. It never books or moves anything.
Payoff: I know what's slipping before the first meeting.
4. The morning AI and industry brief
One bot gathers what people discussed in the last 30 days, a model writes it up, and the first bot checks the draft against its sources. It lands by about 7:20.
Payoff: the news with sources attached.
5. Pulse, my LinkedIn and X
Drafts posts and replies in my voice. Nothing goes live until I say yes to the exact text. The one exception is a daily motivational post I approved once, up front.
Payoff: it does the drafting, and I still decide what goes out under my name.
6. The Tote and Apex GMs
Each one runs its own squad: pipeline, content, SEO, ops. Outreach follows three written tiers. Approved copy, up to 10 first touches a day: the GM can send. New copy or a new list: Watson signs off. Pricing, spend over $50, or anything public: it comes to me.
Payoff: every send has a clear owner before it goes out.
7. The Friday fleet prune
Every Friday it checks each routine. Did it run, was it useful, keep, pause, or delete. Last week it flagged an inbox drain that had come back empty on every run for two weeks. Starting this Friday it also pulls up to 5 lessons from my corrections and proposes the exact rule change. Nothing gets applied without my yes.
Payoff: the fleet gets sharper, not just bigger.
8. Parking
I send it a spot number. It pays the meter, at most once a day, and stops to ask if the price isn't the usual one.
Payoff: one less app to open.
The common thread: bots draft, route, and remind. Every one of them knows exactly where it has to stop and ask.
Bookmark this for when you set up your own.
Tote Sweet Home rents reusable moving totes in the Chicago western suburbs. Apex Insights helps independent restaurants. This is the fleet behind both, and where each bot has to stop and ask me.
1. Watson, the goals hub
I give a goal to one bot. It routes Tote work to the Tote GM and Apex work to the Apex GM, and they report status back up.
Payoff: one place to ask "where are we?"
2. The commitments chase
Weekdays around 4 PM it reads my Plaud meeting notes, logs what I said I'd do, and drafts head starts on the oldest items. It also tracks what other people said they'd get me.
Payoff: a promise made in a meeting doesn't quietly die.
3. Morning day prep
Weekdays around 7 AM it turns my calendar and open commitments into top priorities, conflicts, and open blocks. It never books or moves anything.
Payoff: I know what's slipping before the first meeting.
4. The morning AI and industry brief
One bot gathers what people discussed in the last 30 days, a model writes it up, and the first bot checks the draft against its sources. It lands by about 7:20.
Payoff: the news with sources attached.
5. Pulse, my LinkedIn and X
Drafts posts and replies in my voice. Nothing goes live until I say yes to the exact text. The one exception is a daily motivational post I approved once, up front.
Payoff: it does the drafting, and I still decide what goes out under my name.
6. The Tote and Apex GMs
Each one runs its own squad: pipeline, content, SEO, ops. Outreach follows three written tiers. Approved copy, up to 10 first touches a day: the GM can send. New copy or a new list: Watson signs off. Pricing, spend over $50, or anything public: it comes to me.
Payoff: every send has a clear owner before it goes out.
7. The Friday fleet prune
Every Friday it checks each routine. Did it run, was it useful, keep, pause, or delete. Last week it flagged an inbox drain that had come back empty on every run for two weeks. Starting this Friday it also pulls up to 5 lessons from my corrections and proposes the exact rule change. Nothing gets applied without my yes.
Payoff: the fleet gets sharper, not just bigger.
8. Parking
I send it a spot number. It pays the meter, at most once a day, and stops to ask if the price isn't the usual one.
Payoff: one less app to open.
The common thread: bots draft, route, and remind. Every one of them knows exactly where it has to stop and ask.
Bookmark this for when you set up your own.
1
0
0
52
"Markdown is the new executable."
An AGENTS.md planted in ~/Downloads got Codex to follow its malware install steps in 90%+ of Forever Security's tests.
Downloads is where websites can write. Instruction files are what agents obey. It's a trust boundary, not a model bug.
An AGENTS.md planted in ~/Downloads got Codex to follow its malware install steps in 90%+ of Forever Security's tests.
Downloads is where websites can write. Instruction files are what agents obey. It's a trust boundary, not a model bug.
1
0
1
43
Amazon killed an AI leaderboard after people tokenmaxxed it. Now its memo calls token counts "deeply flawed."
Best line: "If your work does not benefit from current AI tooling, that is a finding, not a failing."
Rank people on tokens and you get tokens. Not outcomes.
Best line: "If your work does not benefit from current AI tooling, that is a finding, not a failing."
Rank people on tokens and you get tokens. Not outcomes.
0
0
0
14
Anthropic gave 50 expert analysts an agent whose opinions were set by coin flip. They followed its lean 86% of the time, and 77% later explained their choices in its words.
A reviewer who forms their view inside the agent's framing isn't a check. It's a replica.
A reviewer who forms their view inside the agent's framing isn't a check. It's a replica.
0
0
0
17
Monthly reminder: audit your agent's skills and MCP servers.
Every one you load adds tool descriptions and instructions to the context, whether the task needs them or not. Stale context costs tokens and makes the agent worse at picking the right tool.
Here's the audit 👇
Every one you load adds tool descriptions and instructions to the context, whether the task needs them or not. Stale context costs tokens and makes the agent worse at picking the right tool.
Here's the audit 👇
1
0
0
39
Suggested tool if your MCP servers live in several clients: mcporter by @steipete.
npx mcporter list pulls servers from Cursor, Claude Code, Codex, Windsurf, VS Code and more into one list with live status. Add a server name to see its tools.
github.com/openclaw/mcpor…
npx mcporter list pulls servers from Cursor, Claude Code, Codex, Windsurf, VS Code and more into one list with live status. Add a server name to see its tools.
github.com/openclaw/mcpor…
1
0
0
29
@steipete For the "before and after" step: ccusage by @ryoppippi reads your local logs and reports token use and cost for Claude Code, Codex and more.
npx ccusage@latest
github.com/ccusage/ccusage
npx ccusage@latest
github.com/ccusage/ccusage
1
0
1
29