26 / ai engineer / build agent setups guides on grok bot + claude · ads via @CELE8EX https://xitter.gq/t.co/CjJFiW5CwC
Brooklyn, NYC
Joined October 2023
- Tweets28.1K
- Following276
- Followers3.2K
- Likes52K
bobaaa retweeted
Can your AI agent survive its own death? On Lizard it keeps up to 500 GB of files after its sandbox dies
lizard.build/volume
Lizard Persistent Volumes are NVMe SSDs that live outside the sandbox:
> plug one in and your agent saves its files to it
> kill the sandbox, the disk and the files stay
> start a new sandbox with the same disk, and the agent picks up where it stopped
> you pay only for the space you use, $0.14 per GB a month
bobaaa retweeted
SCREEN STUDIO IS $29/MO. I REBUILT ITS AUTO-ZOOM IN ONE NIGHT, FREE
that's $348 a year on the monthly plan. mac only, and their FAQ says no near-future plans for windows.
mine runs in a browser tab on any laptop with Chrome. Claude Code + Opus 5.5.
-> glidecut.vercel.app
free & open-source:
> zooms where you click and type
> in 9:16 the frame follows the action
> drag or delete zooms on a timeline
> MP4 1080p 60fps, with sound
> no install, no account, no upload
top is the raw recording, bottom is glidecut, 0 manual edits
first version zoomed 7 times in 22 s. fixed now, code ↓
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
SCREEN STUDIO IS $29/MO. I REBUILT ITS AUTO-ZOOM IN ONE NIGHT, FREE
that's $348 a year on the monthly plan. mac only, and their FAQ says no near-future plans for windows.
mine runs in a browser tab on any laptop with Chrome. Claude Code + Opus 5.5.
-> glidecut.vercel.app
free & open-source:
> zooms where you click and type
> in 9:16 the frame follows the action
> drag or delete zooms on a timeline
> MP4 1080p 60fps, with sound
> no install, no account, no upload
top is the raw recording, bottom is glidecut, 0 manual edits
first version zoomed 7 times in 22 s. fixed now, code ↓
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
bobaaa retweeted
$61M TRIED TO GET INTO A $3.9M TOKEN SALE
jup.ag/tokens/FjPKUGnF5zBpdE…
it's 1win's token. yes, the betting site. official listing, trading on solana via jupiter since monday night
15.64x oversubscribed, so most of that $61M got sent back. whoever got filled paid $0.03
it wicked to ~$0.091 in the first hours, 3x the sale. now it sits around $0.052, still ~1.7x, with ~5,700 holders and $3.5M traded in the last 24h
1win is an actual betting business with revenue, which is probably why the sale got that much. watching whether volume holds once the launch crowd moves on
$61M TRIED TO GET INTO A $3.9M TOKEN SALE
jup.ag/tokens/FjPKUGnF5zBpdE…
it's 1win's token. yes, the betting site. official listing, trading on solana via jupiter since monday night
15.64x oversubscribed, so most of that $61M got sent back. whoever got filled paid $0.03
it wicked to ~$0.091 in the first hours, 3x the sale. now it sits around $0.052, still ~1.7x, with ~5,700 holders and $3.5M traded in the last 24h
1win is an actual betting business with revenue, which is probably why the sale got that much. watching whether volume holds once the launch crowd moves on
bobaaa retweeted
THIS IS F*CKED UP.. YOU SPEND MONDAY IN CURSOR, TUESDAY IN THE GROK APP, AND ON WEDNESDAY YOUR GROK BOTS JUST STOP
nothing broke. on the X subscription it's all one tank. the docs say grok bot "draws from the same weekly usage pool as Cursor and the Grok apps"
and the bots spend from it on their own:
> every reply counts, even in group chats and between your own bots
> the Test button runs the real job, so it spends real usage
> tagging your bot on X starts a task on its cloud computer
you can't drop to a cheaper model and limp to friday either, there's no model picker
bookmark this before your next coding session. 7 fixes with the exact prompts in the guide ↓
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
bobaaa retweeted
voice agents keep hearing themselves talk
the agent's own TTS leaks into the mic, STT mistakes it for the user, and the agent interrupts itself to ask ""sorry, could you repeat that?""
@soniox_ai STT v6 tackles this inside speech recognition:
> you stream the agent's outgoing text into STT
> the model knows what the agent just said and filters its own voice out
> it still hears the user when they cut in mid-sentence
the same context helps with short answers, names and numbers, the stuff phone bots often get wrong
and it supports all of this across 60+ languages
Introducing Soniox STT v6, the world’s first speech recognition model that knows what your voice agent is saying and automatically filters out the agent’s own voice.
Two connected innovations make this possible:
→ Agent text input gives recognition context from both sides of the conversation, improving its understanding of user responses.
→ Agent voice echo cancellation filters out the agent’s playback while continuing to listen to the user, even when they interrupt.
The result: fewer false interruptions, fewer clarification loops, and more reliable conversations.
V6 also delivers significantly higher accuracy across 60+ languages for voice typing, meeting transcription, multilingual recognition, and real-time translation, with stronger speaker diarization, language identification, endpoint detection, and context customization.
One unified speech model. All 60+ languages. The full feature set, with languages freely mixed.
Available now for real-time streaming and asynchronous transcription.
Explore what’s new in Soniox STT v6:
soniox.com/blog/soniox-stt-v…
voice agents keep hearing themselves talk
the agent's own TTS leaks into the mic, STT mistakes it for the user, and the agent interrupts itself to ask ""sorry, could you repeat that?""
@soniox_ai STT v6 tackles this inside speech recognition:
> you stream the agent's outgoing text into STT
> the model knows what the agent just said and filters its own voice out
> it still hears the user when they cut in mid-sentence
the same context helps with short answers, names and numbers, the stuff phone bots often get wrong
and it supports all of this across 60+ languages
Introducing Soniox STT v6, the world’s first speech recognition model that knows what your voice agent is saying and automatically filters out the agent’s own voice.
Two connected innovations make this possible:
→ Agent text input gives recognition context from both sides of the conversation, improving its understanding of user responses.
→ Agent voice echo cancellation filters out the agent’s playback while continuing to listen to the user, even when they interrupt.
The result: fewer false interruptions, fewer clarification loops, and more reliable conversations.
V6 also delivers significantly higher accuracy across 60+ languages for voice typing, meeting transcription, multilingual recognition, and real-time translation, with stronger speaker diarization, language identification, endpoint detection, and context customization.
One unified speech model. All 60+ languages. The full feature set, with languages freely mixed.
Available now for real-time streaming and asynchronous transcription.
Explore what’s new in Soniox STT v6:
soniox.com/blog/soniox-stt-v…
THIS IS F*CKED UP.. YOU SPEND MONDAY IN CURSOR, TUESDAY IN THE GROK APP, AND ON WEDNESDAY YOUR GROK BOTS JUST STOP
nothing broke. on the X subscription it's all one tank. the docs say grok bot "draws from the same weekly usage pool as Cursor and the Grok apps"
and the bots spend from it on their own:
> every reply counts, even in group chats and between your own bots
> the Test button runs the real job, so it spends real usage
> tagging your bot on X starts a task on its cloud computer
you can't drop to a cheaper model and limp to friday either, there's no model picker
bookmark this before your next coding session. 7 fixes with the exact prompts in the guide ↓
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
bobaaa retweeted
THIS PERP DEX IS HANDING TRADERS 32% OF ITS TOKEN
it's variational. VAR tge is planned for q4, so ~2 months of points left
your share = your points ÷ all points × 32% × fdv. with ~10.7M points out, that's ~$30 a point at $1B fdv, ~$60 at $2B. on paper, the fdv is a guess until tge
points cost money to farm. no maker/taker fees, but you pay spread, funding and slippage, and a low fdv can turn the farm into a loss
cost per point and the hedge setup are in the article below
omni.variational.io/?ref=OMN…
THIS PERP DEX IS HANDING TRADERS 32% OF ITS TOKEN
it's variational. VAR tge is planned for q4, so ~2 months of points left
your share = your points ÷ all points × 32% × fdv. with ~10.7M points out, that's ~$30 a point at $1B fdv, ~$60 at $2B. on paper, the fdv is a guess until tge
points cost money to farm. no maker/taker fees, but you pay spread, funding and slippage, and a low fdv can turn the farm into a loss
cost per point and the hedge setup are in the article below
omni.variational.io/?ref=OMN…
bobaaa retweeted
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
bobaaa retweeted
This is f*cking gold.. ONE GROK BOT ROUTINE ON A 10-MINUTE SCHEDULE RUNS 1,008 TIMES A WEEK. MOVE IT TO 3 TIMES A WEEKDAY AND IT'S 15. SAME DIGEST, 67X FEWER RUNS.
this is a full grok bot usage guide. every line in the official docs that spends your week, and the exact message to send your bot for each fix.
> usage counts agent steps and tokens, not messages
> every reply between bots spends your week too
> the test button does the real work, so it costs a real run
> on the x subscription, cursor and the grok apps eat from the same week
honestly my favourite line is from the docs: "a slack trigger on a busy channel can use a week of usage in a day."
i kept the boring job for myself: saying ok before the bot pauses anything.
how the week runs out on day 2 is on screen in the clip.
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
GROK BOT HIT 100% OF MY WEEKLY LIMIT ON DAY 2
a routine every 10 min runs 1,008 times a week. bot-to-bot messages, the Test button and @bot tags spend it too
7 fixes with the exact prompts. bookmark before your meter resets ↓
DEEPSEEK OPEN-SOURCED THE AGENT THAT GREW OUT OF ITS OWN TEAM'S DAILY WORK. 12,000 STARS IN TWO DAYS
you mention it in a github issue. it opens a task, works in its own session and folder, and replies in the same thread with a pull request
it's designed to take as little of your attention as possible:
→ it acts only when someone you allowed asks, or a team rule says so
→ a web panel shows only what needs a human decision
→ your team chat can call it too
setup is a conversation with your own coding agent:
"Read SETUP.md in this repository and set up dsh-superagent with me"
MIT. deepseek calls it a research prototype, not a product
repo in the first reply
ONE SENTENCE IN. ARCHITECTURE DIAGRAM OUT
a new claude skill draws your whole stack from one sentence. official logos, no arrows through text, light and dark
you type:
"draw our docs chatbot: next.js on vercel searches supabase pgvector and answers with claude. inngest syncs notion every night."
you get back:
→ one html file with fonts and logos inlined, opens offline
→ png in both themes
→ a react component with zero deps
→ or a claude artifact you share as a link
how it stays clean:
→ logos come from svgl, devicon or the brand's own repo, never recolored
→ aws stacks get aws's own icon set
→ lines route around labels, and the agent renders the diagram and checks it before handing it over
works in claude code, codex and cursor. MIT, 880+ stars in its first two days
install:
npx skills add caru-ini/stack-diagram
bookmark it for your next readme or pitch deck
ELON JUST ADMITTED GROK BOT COSTS TOO MUCH TO RUN. PEOPLE ARE HITTING 100% USAGE DAYS BEFORE RESET
his words, nov 1: "the cost of serving grok bot was unsustainably high relative to revenue"
so the weekly limit is the thing to protect now. 6 fixes that make it last:
→ ask your bot for a usage dashboard: each bot, how often it runs, its share of usage, and whether you opened its output
→ routines no more often than hourly. a news bot checking every 10 minutes is the usual leak
→ filter events before the bot wakes up, not after
→ anything you ask twice becomes a script or a skill. the bot writes it once
→ pause bots whose output you never read
→ load tools on demand instead of every connector on every run
the rule: pay only for output you actually read
full dashboard prompt in the first reply