Skip to content

Aye Robot comic (@ayerobotcomic)

Field Value
Status Concluded — experiment complete (2026-05-03). Gate pattern proven end-to-end: brainstorm → Telegram approve → xurl post to @ayerobotcomic. Stopping intentionally: idea quality drifted, growth path is manual-only (auto-like/follow removed from API), and the experiment served its purpose as Waypoint 1 proof-of-concept. Learnings (gate pattern, xurl transport, Crusty cron, Telegram approval loop) carry directly into future Operator Agent ops work.
Type Daily one-panel comic (X / Twitter)
Unblock X API pay-per-use (2026-04-17) removed the economic barrier; 2026-04-21 config/skill fixes removed the OpenClaw integration blockers — detail in projects/aye-robot-crusty-paused-x-automation.

Description

Public daily cartoon on a dedicated X account: https://x.com/ayerobotcomic. One panel per day; strips are HN-flavored in the same spirit as the Aye Robot idea loop that crusty runs (see glossary). This page is the canonical wiki home for the published comic channel — automation and agent detail stay on crusty and openclaw-autonomy-org.

Autonomous daily loop (Crusty) — gate unlocked (2026-04-27); economics note (2026-04-29): Posting via consumer web + browser automation remains off (policy). API path is live (grok-imagine; xurl). grok-4.20 two-tier routing rolled back toward grok-4-1-fast-reasoning for token burn — projects/crusty Steady-state economics. Morning image → Telegram cron tweaked after intermittent image delivery. Subjective: early ideas > recent batch — still deciding sustained investment. Listen path + post reliability unchanged as next engineering items.

Mini experiment — Crusty vs Grok scheduled task (2026-05-10)

Active (informal): Run the same Aye Robot creative lane in parallel on (A) Crusty — OpenClaw scheduled job / pipeline (grok-imagine, session logs, Telegram delivery where applicable) — and (B) personal Grok — grok.com scheduled task for the comic (ideas + / or daily panel output per task prompt).

Hypothesis space (what to compare):

Dimension Crusty path Grok scheduled-task path
Creative quality Strip concepts + images under Crusty’s prompts / skill chain Same wiki-linked brief + task instructions on Grok
Reliability Cron / scheduler behavior; historically intermittent image→Telegram Scheduled-task consistency (fires, completeness of output, retries)
Audit Session JSONL / OpenClaw logs Grok task history / thread artifacts

Framing: Waypoint 1 is concluded (see status table at top); this is not a reopening of the autonomous growth experiment — it is a short apples-to-apples check on whether browser Grok tasks can match or beat Crusty on output quality and delivery reliability for the comic slice only. 2026-05-20: Crusty harness is now Hermes (OpenClaw cutover). 2026-06-06: Aye Robot daily post working on Hermes — monitoring + adjusting for quality under Hermes scheduling/skills (Telegram + xurl + grok-imagine). Compare against grok.com task path when both are firing. Posting to X still flows through xurl on Crusty when automated posting matters; Grok personal has no sanctioned X-post connector — compare generation + delivery unless/until that changes.

Results: Log subjective winners per dimension here when there is enough sample (dates + notes). Open until filled.

Next concrete step — mentions listen path

The next build is the listen half of the loop (the post path is proven). In parallel, harden the same post pipeline for daily reliability (on the order of one week of muscle memory on Crusty — schedule + prompt + session-log checks).

  1. Scheduled read of @ayerobotcomic mentions on Crusty: xurl and/or X API v2 GET /2/users/{id}/mentions (resolve the account user id once). Hourly cadence matches the ~$0.72/mo mentions-poll line in the wiki cost model (720 owned-reads/mo) — see projects/aye-robot-crusty-paused-x-automation Pricing / table.
  2. Telegram rollup — digest of new @mentions to Darcy (expand later to quotes / reply threads as needed). Fold in post engagement summaries (below) in the same digest or a separate short rollup.
  3. Optional same pattern as post — draft reply in Telegram; send only after approval (xurl reply, on the order of ~$0.01 per summoned reply in the pay-per-use model — same doc).
  4. Out of scope — X website scripting; auto-like / follow / quote-post (unavailable on self-serve API).

Why this is next (mentions) — after the 2026-04-17 API change, growth is content + human-gated replies to @mentions, not outreach automation.

Post engagement metrics poll

Goal: see how your own posts are landing (likes, replies, reposts, quotes) without living in the X app — same owned-read pricing as mentions (~$0.001/request after the 24h dedup window; batch tweet IDs per call).

  1. Collect tweet ids for a rolling window (e.g. last 7–14 comic posts, or since last digest) from GET /2/users/{id}/tweets or a cached list updated when you post.
  2. Poll public metrics — GET /2/tweets with tweet.fields=public_metrics (and ids comma-separated) to read like / reply / retweet / quote counts.
  3. Cadence — e.g. 2–4×/day for fresher signal, or once daily for a lighter bill; tune against projects/aye-robot-crusty-paused-x-automation cost table.
  4. Telegram — short table or delta vs last poll (optional) alongside the mentions digest.

Why metrics — mentions catch @ayerobotcomic; metrics polling catches silent engagement and helps you spot what resonates before you iterate captions or timing. Listening = mentions plus optional metrics on your timeline.

Workflow (current)

Step What happens
Ideas Grok task / Crusty can propose ideas; image side is on a morning schedule to Telegram (2026-04-27).
Pick Human chooses the best concept.
Produce Image via Crusty’s workspace grok-imagine skill (xAI grok-imagine-image) — not OpenClaw’s generic image_generate tool. Grok daily task remains a parallel option. Exec ground truth: session-log audit trail (recurring-issues, projects/aye-robot-crusty-paused-x-automation). Manual touch-ups before post (optional).
Ship X API v2 via xurl (live path proven 2026-04-27 after Telegram approval). Browser-driven X is not used.

Workflow (target — Crusty / Grok-only)

Single OpenClaw run on Crusty's xAI / Grok-only stack (no other LLM API on the machine), in order:

Step Target behavior
Ideas Crusty generates strip concepts (LLM / tools).
Pick Crusty or human selects the best.
Image grok-imagine workspace skill (single shell command; no compound cd &&).
Post X API v2 via xurl: xurl post "…" --media … on @ayerobotcomic (OpenClaw's merged skills/xurl/SKILL.md).
Listen Next concrete step — mentions + optional engagement metrics → Telegram digest → optional gated xurl reply.

Sequencing (2026-04-27): Image + xurl + Telegram gate proven end-to-end. In flight: (1) Listen path — @mentions + optional post metrics → Telegram digest → optional xurl reply, (2) post path reliability (same brainstorm → approve → xurl pipeline, ~1 week to steady daily). → projects/crusty (Waypoint 1), projects/aye-robot-crusty-paused-x-automation.

Cost & commitment

  • Dedicated X account runs about $8/month (premium tier as of setup).
  • Duration is intentionally open — treat as an active experiment; scale or sunset without implying a long-term promise in wiki tone.

Traction (early)

  • 2026-04-12: Four consecutive daily posts (streak); reach still small (typical for a new handle).
  • First posts were manual polish + publish; wiki tracks the project as live and active regardless.
  • 2026-04-17: Grok daily task for the panel image is steady (daily free output); bottleneck shifting to X distribution (automate @ayerobotcomic posting).
  • 2026-04-21: account at 7 posts / 1 follower — recommendation graph barely knows the handle exists.

Distribution & growth (manual, 2026-04-21)

Constraint: auto-follow / auto-like / auto-quote removed from self-serve X API (same pricing change as projects/aye-robot-crusty-paused-x-automation); browser-scripting X = suspension risk per X automation rules. Growth is manual: cadence + targeted follow + hand-gated replies.

Highest-leverage moves (ranked)

  1. Ship daily. Consistent cadence teaches the For You graph what this account is — live post path unlocked (2026-04-27); harden that path ~1 week + add the mentions listen path.
  2. Same-day quote-reply to big AI news with a relevant panel — best free-impressions lane left. Needs Crusty loop fast enough to turn a news event into a panel the same day.
  3. Manual targeted follow, ~10–15/week — follow-notification ping + seeds the recommendation graph. Tiers below.
  4. Hand-gated replies in threads of Tier 1/2 accounts when the day's panel is on-topic.

Who to follow (tiers, follow-back odds descending)

  1. Comic-adjacent tech humorists — peer group, highest follow-back rate (e.g., @smdiehl, @vboykis, @swyx, @simonw, @willmcgugan, @hmason, @goodside).
  2. AI-twitter humor accounts (personal, not brand) — audience overlap near-total (e.g., @tszzl / roon, @repligate, @nearcyan, @teortaxesTex, @literallydenis, @altryne).
  3. HN-culture operators — match the HN-flavored positioning (e.g., @paulg, @patio11, @dhh, @antirez, @ID_AA_Carmack, @garrytan).
  4. AI researchers who meme — high-reach, selective follow-back but strong recommendation-graph signal (e.g., @karpathy, @jeremyphoward, @hardmaru, @DrJimFan, @emollick, @_jasonwei, @cHHillee, @soumithchintala).
  5. Robotics / autonomy individuals — the about-robots joke audience (e.g., @adcock_brett, @chris_j_paxton, @drfeifei, @pabbeel, @svlevine).
  6. Daily / one-panel comic accounts — near-100% follow-back, overlapping audience (e.g., @xkcd, @ZachWeiner, @sarahcandersen, @nathanwpyle, @MrLovenstein, @poorlydrawnlines, @falseknees).

Anti-patterns

  • Mass-follow of generic "AI" aggregators / course-sellers / tool brand handles — no engagement, no follow-back, pollutes For You.
  • Any scripted follow / like / reply → policy violation → permanent suspension risk.
  • Following AI lab brand handles (@OpenAI, @AnthropicAI, @GoogleDeepMind, @xai) — fine for your own feed but low-signal for their graph noticing you; prioritize individuals inside those labs.

Marketing — “for the machines” (bot-savvy audience, 2026)

Positioning: Lean into HN / AI / robotics meme overlap: the comic is literally Aye Robot — “market to bots” (bot accounts, AI builders, autonomy enthusiasts) is an on-brand differentiator, not a dodge. X’s rules (through 2026) care about behavior (spam, coordinated inauthentic amplification, unsolicited automated engagement, repetitive identical content), not whether your audience includes bot or AI-curious accounts.

Safe mental model: Official API + original Grok-generated panels + human-owned account + Telegram-gated replies + mostly manual follow/engagement per this page = aligned with allowed uses (e.g. automatically sharing creative content). Avoid anything that looks like metric gaming or a bot army.

Execution

Area Direction
Branding Bio along the lines of: One-panel comics for the machines — transparent, playful (e.g. note Grok-powered daily drop, Crusty stack if you want builder cred).
Content Meta jokes: bot life, autonomy, prompt fails, the algorithm, “toaster overlords,” etc. X Polls (Humans vs bots — who gets this?) fit the angle. Vary strips; tie to news; recurring characters beat copy-paste.
Growth Still manual — prioritize Tier 2 (AI humor), Tier 5 (robotics / autonomy), and adjacent bot-dev / agent voices already in your orbit. Organic replies in those threads when the panel fits stay in policy if they’re not automated spam.
Meta story Publicly frame the project as a proof-of-concept: autonomous agent runs the loop 24/7 (with human approval where it matters). Builders and bot operators are a natural amplify/follow segment.

Hard avoids (stay clean)

  • Mass follow or auto-engage with bot networks.
  • Coordinated behavior across multiple accounts.
  • Anything that reads as manufacturing engagement or gaming metrics.
  • Low-effort repetition — same joke structure every day without variation.

Related: Post engagement metrics poll for feedback without spammy automation.

Tech stack

Raw sources

  • (None in-repo yet.) Drop a brief under raw/ if you want an immutable snapshot of URLs or workflow.

Known issues

  • Mentions + engagement listen — spec’d as next concrete step (2026-04-27): @mentions poll + optional post metrics poll → Telegram digest; not wired until Crusty runs the jobs. Post path proven same day. → projects/aye-robot-crusty-paused-x-automation
  • Transport contract: X API v2 via X's official xurl CLI + OpenClaw's merged skills/xurl/SKILL.md — live; OAuth 2.0 PKCE via xurl; dev app must be Pay-per-use + Production or calls return client-forbidden.
  • Growth patterns killed by same pricing change: auto-like, auto-follow, quote-post are removed from self-serve API. Growth has to come from original-content-attracts-mentions + manual-gated replies — no outreach engagement via API.
  • Browser automation = permanently off the table (not just paused). Per X automation rules, "scripting the X website" is listed under Don't and can result in permanent account suspension. The current design (daily image post via API + human-gated mentions replies) is explicitly inside X's Do examples — full policy mapping on projects/aye-robot-crusty-paused-x-automation under Policy posture.