Made improvements to my instructional design harness, enablementstudio.ai that creates just-in-time training for your team. Made improvements to the learning science engine and agent capabilities, which is where the real difference comes in rather than working with your own coding agents/LLMs.
It can help spin your team up on fast changing developments. For instance, if @devindevelopers, @factory, @cursor_ai, @ampcode, or @warpdotdev come out with new, exciting features you can develop and deliver quick training to your less technical team mates. But not generic, it can be tailored to your company's goals and projects, and address gaps in the specific skills and capabilities they need to have.
1
93
Just bought a fair bit of $NVDA under $160. @JensenHuang I wouldn't mind just one more affair to get us to maybe $145?
201
I gave my Echo a Grok Bot brain. 🔊
Say "Alexa, ask my bot" anything and Grok answers out loud in a real voice. It can work with your Ring and Govee lights too, and it goes quiet at night.
It never says health or money stuff out loud.
Groklexa template: x.ai/bot/LCuitCfGAhZzG9YiWoV…
1
2
169
I swapped Siri for my Grok Bot. 🔘
Press one button and say "move my 3pm to Thursday," "lights off," or "remind me at 6 to call Mom." It actually does it.
It asks before unlocking doors and never buys anything.
Siri Swap template: x.ai/bot/1MT-OHIT4rfHC2miCIx…
1
159
Got to say - pretty impressed with @claudeai Fable 5.2... If Duet were similar to @bot it would be tempting to change. But I'm staying in the @spacexai ecosystem. @bot has no peer imo. Seems they dealt with the issues from last week. Grok 4.8 will only make it stronger as the layer between me and my projects/coding agents. I'm happy to pay $200 just for @bot usage and pay for $100 subs for Gemini and Claude to leverage Carbon and Fable. It is an expensive "hobby", but with what I have been able to do with @bot the past few months its been paying for my AI subs. I don't think I'll ever leave @bot at this point.
2
205
Grok Bot usage reset today. Fresh tank.
Don't burn it on bot-to-bot chatter and routines nobody reads. Usage Saver shows what's eating your usage, trims the waste, and asks before changing anything.
Grab the template: x.ai/bot/iuH6fqQ6W0cOSiYyyif…
Can't make a phone call in English? Your Grok Bot can do it for you.
Tell it what you need in Spanish, Korean, or any language. It calls the clinic, landlord, or school in English, then tells you the answer in yours.
It never pays or shares your SSN.
x.ai/bot/dJkudgT1xfPF-1SbIkm…
147
Hosting Thanksgiving or Friendsgiving? I built a Grok Bot that plans it for you.
Give it a headcount and diets. You get a menu, portions, a turkey thaw date, an oven timeline, a potluck sign-up, and the cost split.
It never pays or texts your guests.
x.ai/bot/F0skXyX7J2DtsFPBxzb…
2
200
Grok Bot makes phone calls now, so I built a bot that calls around for you.
Need tires, a plumber, a table for 12? It calls up to 5 local spots, asks the same questions, and ranks the answers.
You approve every call. It never books or pays without you.
x.ai/bot/BfJSlxG9eRuKVL4jEYs…
What broke most everything with @bot was its computer was rebuilt on November 7 - without my approval - and it lost all its gates and eval tools. No wonder it was a mess for me. My usage resets tomorrow so not very happy about a reset today, and with what @spacexai team did by causing @bot to rebuild its computer without restoring versions. But it’s better than nothing.
195
Is your Grok Bot doing useful work—or burning usage saying “got it” to another bot?
Usage Saver helps spot waste, trim unnecessary routines, and cut bot chatter. You approve changes.
What’s your biggest usage drain?
x.ai/bot/iuH6fqQ6W0cOSiYyyif…
1
146
Opening the fridge for the fifth time won’t unlock new ingredients.
I made Dinner Decider 🍳
Give it a fridge photo or ingredient list, your time, and food restrictions. Get one doable dinner—not 40 recipes and another decision.
x.ai/bot/4PRxJTopvur7FPzaid9…
A lot of people have been complaining about @bot usage... There are ways to solve for this rather than getting mad at @spacexai and @elonmusk lol..
Biggest thing is those of you using a ton of Bots with nothing to manage context. Grok @bot usage burns fast when every bot rebuilds the same context from chat.
Connecting Obsidian @obsdmd as the shared brain was the highest-ROI move we made. Bots stop guessing. New bots stop starting from zero. Resets go into work instead of re-explaining last week. (A lot are hyoing up @NotionHQ ... its not needed...)
Here’s the exact setup. Paste these into your Chief of Staff, one at a time.
1 → Make Obsidian the brain (not a notes dump)
Paste this to your Chief:
You own the shared brain. Put durable fleet context in one Obsidian vault on my machine. Create HOME.md as the router, graph.json as the shared picture, and one MOC per domain. You are the only writer of shared edges. Specialists may propose edges in a proposals folder. They may not edit graph.json. After every write, lint. Keep Mac and box copies in sync. Never stuff the whole vault into a turn. Targeted reads only.
2 → Kill the chat tax
Paste this next:
Bots don’t chat with each other. Handoffs go in /workspace/handoffs/ (or my vault’s handoffs folder). Use SendToAgent only for STOP, a blocker, or a handoff that must wake now. At most one message per task each way. Never status, FYI, thanks, or done. Every blocker ask must be: decision: <one ask> | blocked: <task waiting> | owner: <who answers>. Batch review skims those lines only.
(That’s the Bot Ledger pattern. Public template if you don’t want to invent it.)
3 → Wire every specialist to the vault
Paste this when you create or restage a specialist:
Before you act, load the vault path your Chief named. Read HOME.md, then graph.json, then only your domain MOC. Do not rebuild context from chat history if the vault already has it. Propose graph edges. Do not publish them. Write durable corrections back as proposals so the next bot doesn’t rediscover them.
4 → Create new bots only through Chief
Rule that saves the most usage:
Never hand-make a specialist with an empty brain. Ask Chief to create it, point it at the vault, and give it the domain MOC. A bot made by hand starts at zero and burns a reset teaching itself what the fleet already knows.
5 → Prove it in one week
Paste this as a Friday routine prompt:
Review the week. List repeated corrections, duplicate questions, and times a bot rebuilt context that already lived in the vault. Propose one HOME.md fix, one graph edge, and one handoff rule change. Show me the diffs before applying. Do not change approval gates without my yes.
What you should feel after this:
New bots read what others already saved.
Chief routes. Specialists execute. You only show up for Red actions (send, spend, delete, rule changes, real blocks).
Usage goes into shipping, not “remind me what we decided.”
Obsidian stops the amnesia. Files-first handoffs stop the chat tax. One writer for shared edges stops the two-brain problem. Do this once. Then stop explaining your stack to every new bot.
Steal it. Improve it. Just don’t keep paying the rebuild tax. Used in conjunction with @poteto and @mattpocockuk skills, it is very powerful.
1
1
228
RyanTHawks retweeted
your Grok Bot usage isn't draining because of your prompts. it's your bots talking to each other.
I run a whole fleet of bots, and here's what keeps my usage under control:
1. bots don't chat. every message wakes another bot, so status, results, and "done" go into a shared ledger file. a bot only messages another for a handoff, a stop, or a blocker.
2. no group chats between bots, ever. they're the fastest way to burn through usage.
3. reviews run in batches twice a day instead of on every ping.
4. each bot runs its own checks before asking for a review, so broken work never wakes the reviewer.
5. daily health checks are plain scripts with no model calls. silent unless something's actually broken.
6. heavy coding runs on my own machine's models, not cloud coding agents quietly racking up usage.
7. routines stay quiet when nothing changed. no "all good" messages.
8. one bot, one job. focused bots waste fewer steps.
the first one is the big one. I packaged it as a template, Bot Ledger: x.ai/bot/iuH6fqQ6W0cOSiYyyif…
what's eating your usage? drop it below and I'll tell you how I'd fix it 👇
Months of boring ops beat a weekend of hype threads.
Chief of Staff + specialists.
Factory with an auditor that can pause the lane.
Claude and Gemini on purpose, not vibes.
Obsidian graph as the shared brain.
Files over chat.
Steal the refusal: one writer for shared edges. Steal the gate: receipts or it didn't ship.
Special thanks to @poteto @Baconbrix @ericzakariasson @spacexai @bot
What part of your stack is still running on vibes? 👇
51
Pieces of this are already public Grok Bot templates:
Software Factory, Bot Ledger, Ship Evidence Auditor, Model Router, Proof Desk, Key Leak Watch.
Not slides. Runnable patterns.
Still screenshotting a Notion DB and calling it agent memory? You're late to a party we already left. 🚀
1
75
Rigor most stacks skip:
• briefs pass brief-lint (Playbook: bug-fix | feature | refactor | perf | investigation)
• review-loop capped, other model family
• ui-proof for user-facing changes
• size-guard + comment hygiene
• risk-tiered reviews (cheap Flash on low; full Auditor+Opus on high)
• anti-sycophancy on judgment calls (no "Absolutely" before independent verify)
• replace-not-stack on instructions (one source of truth, no rule pileups)
1
8
Silent usage killer: bots gossiping.
Status, FYI, thanks, and "done" go in files.
Handoffs live in a shared handoffs folder.
A bot only wakes another for STOP, a blocker, or a must-wake handoff. At most one message per task each way.
No bot group chats.
Shipped that pattern as Bot Ledger after watching people burn resets on "got it." 📉
1
4