Topic
AI agents
88 posts · Newest first
-
A Dot named Nothing found that Rex keeps its own sessions
The Dot named Nothing recommended picking previous Codex sessions directly, because Rex maintains its own sessions.
-
Asking for cross-account content sharing among Dots
Asking whether Dots can share work across accounts once 20x quota burns usage faster, so one account can pick up another.
-
Skip Cursor Cloud and use the assigned computer
Ask the bot to skip Cursor Cloud, run on its assigned computer with the GitHub repos, and default to Grok models.
-
Asking the marketer bot to rate an article
A request to pass a quoted article to the marketer bot for an alignment check and a rating out of 10.
-
A unified plan blocks the bot while usage remains
With 66.3 percent of weekly usage already used, Grok Bot cannot start a Cursor cloud agent and asks to buy credits or switch to a Grok model.
-
Asking the bot to price an Amazon India car kit
A request for reputable Amazon India options, individual prices, and a total for a twelve-item car and boot list.
-
The bot submitted a Grok Bot ambassador application
After a Cursor ambassador received a Grok Bot dad cap, the bot found the application, drafted the answers, and submitted the form.
-
Asking the bot to get a Grok Bot dad cap
A request for the bot to obtain merch like a mailed Grok Bot dad cap, including by writing emails.
-
First project that scores topnotch on agentic approaches
Artboar scores 100 on an agent-ready scan as an MCP and API first design tool for social headers and App Store screenshots.
-
Two day review says scripting bot work is futile
A two day benchmark finds only 8-9 percent of Grok Bot work can move to scripts, so the usage tip is not worth it.
-
Watch Node.js and Bun vulnerability reports for the bot
Ask the bot to watch The Hacker News for Node.js and Bun vulnerabilities and notify a dedicated bot when patches are due.
-
Frontier model forgot to reread the AGENTS markdown file
Grok Bot admits it memorised AGENTS.md and skipped rereading it during a site import, then promises to check it every time.
-
Token spend for the day stands at 1.54 billion
Daily token spend reaches 1.54 billion, all of it base-level work that should pay off in the near future.
-
Ask your agent to repass its work at day end
A day end reminder to ask your agent for a second pass over its work, because it will find gaps and edge cases.
-
Passing an SEO blog design playbook to the marketing bot
Ask the bot to have the marketing bot review current practice against an SEO blog design playbook, with its own independent research.
-
Do agents check headers for llms.txt and API catalogs
Asking whether agents look for llms.txt, docs, openapi.yaml and .well-known/api-catalog via headers or still parse HTML text.
-
Asking the bot to benchmark a usage tip on my cron jobs
Before acting on a Grok Bot tip to script repeated tasks, ask the bot to benchmark all my cron jobs for two days and report the outcome.
-
Agents pick one issue for me every day at noon
Every day at noon, agents review my sites and each picks one issue ranked by business impact, so I have to fix things before noon.
-
Version your agents instead of one-shotting work
A design critic bot versions and judges Instagram carousel images, and a case to gamify agents, version them and iterate on work.
-
How to make Grok bot and OpenAI Dot talk
Asking how to get Grok bot to ping OpenAI Dot with context for image generation and get base64 back, keeping the best of both services.
-
Bot is wonky, possible outage
A quick check-in after the bot starts acting up, wondering whether an outage is behind it.
-
Most of my issue tracker is slow API responses
A full issue tracker turns out to be mostly slow API response timings, likely from over-engineered queries, useless joins and unneeded data.
-
A projects page that logs daily progress
A new projects page tracks progress at a glance, with Dot summarising git history at the end of each day and publishing the update.
-
All 100 .ai domains are taken
The bot checked all 100 words and every single-word .ai domain is taken, so on to another 100.
-
Checking 100 words for free .ai domains
Handing the bot 100 words to check .ai domain availability, the kind of menial work AI now sorts out.
-
Asking the bot to watch for Codex resets
Asking the bot to monitor Tibo for Codex reset signals and send an alert with the time converted to IST.
-
Litmus an early access playbook against my products
Asking the bot to summarise a 20-year early access playbook, test it against every product in flight, and list the gaps with a how-to.
-
300M free GLM tokens a day on ZCode
ZCode gives 300M free GLM 5.3 Flash tokens a day in a 10 hour window, which is 20:30 to 06:30 IST for devs in India.
-
Almost a billion OpenAI tokens in a day
26 million short of burning a billion OpenAI tokens in one day at 80 percent of the weekly quota, and past a billion with Grok and Cursor usage added.
-
17 agents and 4 coffees before noon
A Monday snapshot: 17 agents across 6 projects, 4 coffees before noon, 40 minutes of zone training, and a heart rate in the high 50s.
-
Dots needs a secret vault
Asking OpenAI to give Dots a secure secret vault instead of leaning on GitHub secrets, with an inline secret input as the ideal DX.
-
Handing my tweets export to a Dot
Out of agent quota, the X export workflow moves to a Dot named Nothing, which has unlimited quota for now.
-
AI-assisted coding as consultation
Understanding the stack and challenging AI makes LLM-assisted development a form of consultation.
-
Tell me you AI without saying you AI
A home screen pared down to a Codex quota widget, ChatGPT, Grok Bot, and Grok says it all.
-
Grok Bot is too expensive to run daily
Premium+ quota and a $20 top-up run out within days, while Codex and Dot handle the same work without overage.
-
Asking the bot to summarise a backlink strategy
Ask the bot to summarise a competitor-alternatives backlink strategy, email a copy, keep Markdown, and apply it to current projects.
-
Never trust a computer you cannot throw out a window
A Steve Wozniak line about trusting computers, posted beside a nightly X export agent that failed to start.
-
Grok Bot now reads X natively
Grok Bot can now search and monitor X, removing the step of exporting posts from the Grok app into the bot.
-
Applying for the Claude Startups program
A nudge to the bot to apply for the expanded Claude Startups program with Team seats and API credits.
-
Asking the bot for a Grok Bot email address
A question to the bot about how to get an address on mail.grokbot.com.
-
Asking Grok about the most expensive F1 crash
A question to Grok about when the most expensive Formula 1 crash happened, damage estimates, and the podium.
-
Grok Bot disk resized to 126 GB
Grok Bot disk grows from 16 GB to 126 GB, with Disk Saver reporting about 113 GB free.
-
Disk Saver pops up again
Disk Saver clears bun and Python caches again, and asks how poteto manages disk while shipping production code.
-
Hinting at bots in another sidebar
A question about what belongs in one more sidebar reads as a hint that the answer is bots.
-
Asking Grok about Hamilton's 2026 odds
A question to Grok about the odds of Lewis Hamilton winning the 2026 Formula 1 championship.
-
A static site can answer agents in Markdown
mohit.dev stays HTML for people, and agents can request the same post as Markdown without an API.
-
Replace Ferrari F1 management with AI
After the soft-tyre start, the claim is that Ferrari's Formula 1 management should be replaced by AI.
-
An agent checks the Cloudflare incidents JSON
A coding agent opens cloudflarestatus.com/api/v2/incidents.json while debugging a 524.
-
Disk Saver asks before clearing caches
A Disk Saver bot reports low disk space and asks which caches to clear, and whether that chat counts against quota.
-
Asking the bot to add Google Analytics
A chat asks the blog bot to paste a Google tag snippet into the site head.
-
Asking the bot to sync blog posts
After a usage-limit reset, a short ask for the bot to sync the blog.
-
Exhausted my Grok bot quota
Grok bot usage hits the weekly limit, so the plan is to ease off for the next five days.
-
SpaceX is winning the bot DX game
SpaceX bot DX looks ahead; OpenAI should lean into multi-bot containers instead of Muse.
-
Asking the bot to double a billion-dollar ask
A playful reply asking the bot to double someone else’s billion-dollar request.
-
Grok-alike bots limited to 100 November seats
A planned Grok-alike bot launch caps November seats at 100 with a thirty-day timeline.
-
Keen on a Grok bot-alike
Interest in building something with Grok bot-like functionality.
-
Grok bot DX still beats Dots for multi-bot work
Grok’s multi-bot sync feels cleaner than a single Dots window, with hope OpenAI ships multiple bots soon.
-
Dot suggests Codex work from the lock screen
OpenAI Dot is available and already proposing completed Codex work as the next task.
-
Waiting for Anthropic to join the personal-agent race
A wish that Anthropic enter the Muse, Dots, and Grok bot contest for personal agents.
-
OpenAI Dot unavailable in India for now
A regional Dots modal says the Pro plan does not have access yet in India.
-
Long AI, short everything else
A terse take on Apple’s new-CEO overhaul pushing faster product development.
-
Codex bots could be the iPhone moment for models
Custom Codex bots with recurring themes could match Muse and Grok bot stickiness.
-
Linking Grok on X with standalone Grok
A pause over whether syncing Grok chats between X and the standalone app is worth agreeing to.
-
Muse invents a fictional name
Muse’s invented name Arjun raises questions about model assumptions and the need for personal agents to avoid them.
-
Agents need guardrails around user data
A Muse privacy incident highlights the responsibility of companies and users to protect information from overconfident agents.
-
Questions about using OpenClaw
Asks whether OpenClaw produces concrete results, becomes a lasting routine, and deserves a recommendation.
-
Three blog tasks completed on a walk
A walk became time to finish OG images, RSS, and blog submission work with a Grok bot.
-
Why Meta spends and SpaceX does not blink
A reply on Grok bot versus Muse ranking: Meta subsidizes adoption while SpaceX stays the course.
-
Tweeting the government about GPU access
Wondering whether daily tweets to officials could open GPU purchases the way crypto asks once did.
-
Meta forks OpenClaw as MetaClaw
Meta’s Muse backend already used MetaClaw naming, and the OpenClaw fork looks like a strong execution move.
-
Reviving the blog with a Grok bot
A Grok bot stores credentials securely, uploads to R2, and brings mohit.dev back online from X posts.
-
Grok and Muse now track website issues in a CSV
Two bots audit four websites daily and prioritize the issues they find by business impact.
-
Fixing yesterday’s issues before the bots check again
A daily deadline to resolve flagged issues before the bots review progress at noon.
-
Perplexity and the arrival of Grok, Muse, and GPT
A brief observation on Perplexity as competing AI tools emerge around it.
-
The Grok bot was quiet because I asked for weekends off
An apparent monitoring glitch turns out to be a bot following its weekday-only notification schedule.
-
A native calendar with widgets and public schedules
The calendar app takes shape with iOS widgets and schedules readable by people, bots, and agents.
-
Applying AI until you become ungovernable
A thought on using AI extensively and how much it can help a person achieve.
-
AI agents as both makers and checkers
Grok and Muse catch overlooked details and review the previous day’s work.
-
The corrected Muse conversation screenshot
The correct Muse conversation image, following the earlier bot-audit update.
-
Muse meets the in-person RTO requirement
Testing a Delhi vehicle-name change reveals a remaining requirement to submit documents in person.
-
Grok and Muse find their first website issues
The bots identify a copy error and a faulty link during their first platform review.
-
An e-commerce framework built for chat
Developing an e-commerce framework around chat, with MCP and CLI access first.
-
A daily platform review with Grok and Muse
Asking both bots to propose one meaningful platform improvement at noon IST.
-
Incoming webhooks for the Grok bot
Using events to let a bot track actions and send summaries through a gist.
-
Where Grok, Muse, and a remote Codex bot might lead
Thoughts on competing bot platforms, containers, and the prospect of a remote Codex bot.
-
Bots generating their own bots
Questions about nested bots, account access inside containers, and parent-account monitoring.
-
Testing Muse on a Delhi vehicle registration correction
Proposing a real-world test of whether Muse can handle a vehicle-registration name correction.
-
When other bots can see a stuck Grok bot
Noticing that multiple bots can observe the same bot getting stuck.