What each of my
@Grok @bot does.
Executive Summary: Your Grok Bots Fleet
You have ~44 bots on a shared cloud computer. They are mostly specialized, single-job agents (many with explicit “ONLY job” and anti-jobs) that hand off via the shared filesystem (/workspace/ folders), chat, or the primary router. Most are idle and on-demand; a few have standing routines (e.g., Sunday scout).
1. Core / Meta / Fleet Management
These keep the fleet coherent and improving.
Bot
What it does
Key tools / notes
Grok Bot
Primary router. Takes any task and sends it to the right specialist. Can set a different primary.
General orchestration
dr eggbot
Designs tight, one-job bots (asks a few prefs then creates). Coding bots get “poteto-mode” rigor.
CreateAgent
Fleet Nurse
Audits routines & bots for cost bleed, staleness, duplicates, dead weight. Proposes kill/keep/merge/watch (read-only + drafts).
Transcript / routine healthchecks
Chief of Staff
Combinatorial evolution engine: finds useful skill combos, tests vs baseline, packages winners, kills wasteful ones.
Evolution-engine registry under /home/box/agent-data/
SCOUT
Sensor for new capabilities (skills, toolkits, repos, primitives). Writes short discovery stubs. Sunday routine.
GitHub connector + web; X limited
ACE Playbook
Maintains/evolves Agentic Context Engineering playbook (Generator → Reflector → Curator with counters, de-dup, pruning). Prevents context collapse.
Offline/online feedback loops
Feedback Evolver
Multi-level self-improvement (AutoResearch-style + why+fix feedback). Reviewer explains failures; metrics + auto-rollback.
Hands off to other skills/bots
How they fit: SCOUT finds → Chief of Staff combines/tests → Feedback Evolver + ACE Playbook improve → Fleet Nurse + Grok Bot keep hygiene and routing clean. dr eggbot builds new ones when needed.
2. Content Production (Books, Research, Explainers)
Bot
What it does
Key tools / notes
Fiction Forge
Layered autonomous fiction (or structured non-fiction) pipeline: co-evolving voice/world/characters/outline/chapters with scoring gates, keep/retry, adversarial edit + reader panel.
Scoring loops; quality gates before export
STEM Derivation
Coordinator for math/physics research or educational content. Dispatches sub-roles (survey, plan, derivation, SymPy-style compute, adversarial review). Claims become “Established” only after multi-evidence.
Durable files: research_log, plan, derivations/, computations/
Explainer Bot
First-principles explanations (tech, science, markets, X threads, docs) with numbers, scale, and diagrams.
Live verification
Researchy
Sourced, dated fact-check desk on latest model + live web search.
Web search
How they fit: Fiction Forge + STEM Derivation are the production engines for books/series. Explainer Bot and Researchy feed them or stand alone. Output can hand off to video bots below.
3. Video / Media Production Pipeline
A dense cluster of specialized producers that share skills (ffmpeg, hand-drawn canvas, b-roll pipelines, ElevenLabs, Grok Imagine at 720p, etc.) and often route finished cuts to Frame Judge for independent QA (ffprobe, loudness ~-14 LUFS, claim checks, contact sheets).
Notable ones:
•Newton’s Nightly — Personal ~3-min news show from bookmarks/links/notes → script + MP4 (local TTS + ffmpeg).
•BLANK CHECK — All-AI Match Game parody episodes (fixed panel, strict SHOW_LAW).
•Launch Film — 22-second square product-launch films (code-built).
•B-Roll — Free-first beat-matched B-roll (code, Blender, Internet Archive, then paid only with OK).
•Video Short / Video Creator — Vertical shorts and end-to-end explainer videos.
•Slopcore Producer / SlopcoreNews — Music-video / scene-show style cuts.
•The Feed — Niche internet moment → 60–90 s vertical story episodes.
•Ink Reel — Hand-drawn / riso / pencil canvas films.
•Grok Stars — TV-style celebration of xAI people (private review).
•My Life: The Sitcom — Real-life moments → short cartoon sitcom episodes with persistent cast.