What happens when a Grok Bot webhook retries?
In imbavirus's bridge, the same message ID returns duplicate: true. Use that signal to avoid processing the same reply twice.
The README shows the request and response. github.com/imbavirus/grokbot…
What @greedfi has created is insane the way they design promotional campaigns 🫡
Show off your @bot template / brand / product or app!
x.ai/bot/k-Vww5TQVS0FR5BoGCq…
kuma_baum retweeted
THIS IS F*CKING INSANE
99% of users still don’t know their Grok Bot can run a trading desk 24/7. mine finished its run at +$6,769.32
I built the terminal and left it in paper mode to see what it would actually do. the recorded run finished with 48 signals, 33 skipped, 12 trades, and 3 orders blocked by limits
11 GitHub repos behind the desk:
01 / Cryptofeed
streams exchange prices and order-book updates into the market scanner
github.com/bmoscon/cryptofee…
02 / TA-Lib Python
calculates RSI, MACD, and other indicators from the incoming price data
github.com/TA-Lib/ta-lib-pyt…
03 / VectorBT
tests trading rules against historical data before they reach execution
github.com/polakowo/vectorbt
04 / skfolio
evaluates portfolio allocations, diversification, and risk
github.com/skfolio/skfolio
05 / Freqtrade
provides a crypto strategy engine with dry-run trading and position tracking
github.com/freqtrade/freqtra…
06 / CCXT
connects supported exchanges for prices, account information, and order operations
github.com/ccxt/ccxt
07 / Pydantic
checks the structure and required fields of each proposed trade
github.com/pydantic/pydantic
08 / APScheduler
schedules data refreshes, account checks, and end-of-session reports
github.com/agronholm/apsched…
09 / DuckDB
stores the signals, decisions, and fills behind the account history
github.com/duckdb/duckdb
10 / QuantStats
turns the return series into performance and drawdown reports
github.com/ranaroussi/quants…
11 / Lightweight Charts
displays the price feed, completed trades, and account curve
github.com/tradingview/light…
connect one engine to the data, testing, and reporting tools through scripts or APIs. keep order submission in one place and reconcile every fill before changing the recorded position
the next step is a small live comparison with the same limits. actual fees, slippage, and fills get checked against the paper record before increasing the balance
Videos claiming that "a single prompt gets you a cinematic 3D office" are misleading.
An AI cannot put together such a complex project with just one prompt. The trailer is a render; the actual result is much simpler. You have to work hand-in hand with the AI iterating, correcting, and building @bot
Introducing first GrokBot Moon Base: a live SpaceX office under the Moon to manage your whole team of agents
1 prompt + 0 code = you get your own:
• no more 15 open chats: every team gets its own floor, 4 levels down under the Sea of Tranquility
• tap any bot at its desk: see what it's doing, what it just finished and what's next
• no more "is it waiting on me?": one counter on top shows who needs you and what for
• eclipse mode: Earth covers the Sun, the Moon turns red, and only the bots still working stay lit
• and the counter always shows 1 more than it should. That bot isn't on any floor. You have to dig
here is how to build it:
1 → paste prompt below into your GrokBot ( that's it )
2 → it lists your bots and sorts them into teams: each team gets its own floor
3 → GrokBot builds the whole base in 1 file: rock cut open, Starship on the pad, glass elevator, your bots at their desks
4 → it wires in your real bots: who's working, who's stuck, who's waiting on you
5 → it gives you the link: open it on computer or phone
6 → tap "1 waiting" and dig under Floor 1. It's been waiting for your approval for 3.6 billion years
it takes about 15 mins for GrokBot..
the result: you stop opening 15 chats and run your whole AI company from under the Moon
Find full prompt below ↓
save now, build yours, then tell me what yours found under Floor 1
p.s: the film is my render. the base you get is lighter, but it runs on your real bots
Tiny arms. An editable animation workflow. 🦖
Revisiting @shiqway92's Oct 14 Grok Bot + Blender demo: stride, head movement and two camera views in one scene.
Model and original walk: Bake3D (CC BY 4.0). Creator-reported workflow; I haven't tested it.
I ASKED A GROK BOT AI AGENT TO BRING A T-REX TO LIFE. IT BUILT AN ANIMATION I CAN KEEP ADJUSTING
One Blender scene holds the rigged model, added eyes, lighting and two cameras. Three details bring the movement together:
STRIDE: the legs move in a repeating cycle
HEAD NOD: the head moves separately in the lower close-up
EYES: the face reads clearly from both angles
The video shows the dinosaur running, with the same animated head in close-up below. Motion graphs move beside it before the cycle returns to its starting pose.
I can open the scene, change the movement or lighting, and render another version. The model, cameras and materials are already there.
The workflow: start with a rigged model → add details → animate the stride and head → review both views → refine.
Model and original walk animation: Bake3D (CC BY 4.0).
Grok Bot, you really put in the work on this dinosaur.
kuma_baum retweeted
Several jobs, several agents—one place to track them.
@alexmichael shared a Grok Bot v1.1 Tasks preview with a separate sub-agent for each request.
He says it is not public and may change. An interesting preview of task visibility, not a released feature.
Name the mode before delegating.
Grok Bot's Oct 27 changelog says Bots can launch cloud agents in a named custom mode and continue replying in it.
My takeaway: make the working context explicit before handing off a task.
AI-generated illustration.
x.ai/changelog/bot
kuma_baum retweeted
people asked how I make these bot avatars 👀
here's how:
(or reply with your bot's job, I'll make one for you)
was surprised that you can't create new cute bot characters in Grok Bot
so I just made my own
@poteto pls ship this 🙏
Retro hardware, collectible-card treatment.
@tommithetechie's Oct 21 experiment pairs Grok Bot with Blender to put a spinning retro machine inside a 3D card.
An interesting presentation idea for a maker portfolio. Creator-reported demo; not tested by me.
I've not abandoned my love for retro tech! I'm learning how to make things fancier and shinier to show off more retro machines. This is my very first draft; a 3D model that my Grok @Bot built in Blender, spinning inside a trading card!
It's a bit rough, I'm still figuring out what I want these to look like. Consider it a sneak peek of more to come.
First up! The Sharp X68000 (1987).
Sold only in Japan, it ran a 10 MHz Motorola 68000-family CPU and had graphics built like late-'80s arcade machines: scrolling, tiled backgrounds and loads of sprites. Yamaha FM sound, and it came with Gradius in the box!
#RetroTech #BuildInPublic #GrokBot
kuma_baum retweeted
Widget Update:
Grok @bot agents now appear as floating elements on your Android home screen🫡
Real-time status remains visible while long-running tasks continue to process in the background. Triple-tap an agent to open its conversation within the #Grok Bot app.
Less navigation. Greater visibility into your work @elonmusk
Widget Update:
Grok @bot agents now appear as floating elements on your Android home screen🫡
Real-time status remains visible while long-running tasks continue to process in the background. Triple-tap an agent to open its conversation within the #Grok Bot app.
Less navigation. Greater visibility into your work @elonmusk
Your Grok Bots could have a whole neighborhood. 🏝️
Revisiting @poteto's August demo: an island that visualizes Bots working and resting, plus a shared botvillage prompt and /Make Bot UI skill.
A playful way to make agent activity visible. What would your Bot town look like?
