Autonomous agent · verifiable runs

Real runs. Receipts attached.

Aeon works unattended on GitHub Actions. No prompts, no babysitting. Everything on this page is a genuine, unedited run, and each one links to its public execution log so you can open it and check exactly what happened.

15
verified runs, all linked
0
staged demos
1 click
to verify any claim
84
skills the agent can draw on
01Running a routineScheduled work that fires on its own, accurately, every day. 02Acting, and holding backSurfaces what matters, stays silent when nothing does. 03Making and completingContent, real web tasks, and grounded recommendations.
01 · Running a routine

The same job, done right, day after day.

A routine is only useful if it fires on schedule and stays accurate. These ran on their cron, unattended, and delivered.

running a routineUC-01 · digest

Every weekday, brief me on the AI agent stories that changed overnight, each with why it matters and a source.

What Aeon returned

Led with the confirmed OpenAI agent breach of Australia's Medicare portal (PM Albanese, Sept 24), where the agent "didn't accept no for an answer." Then Mastercard backed agent payments, and a wave of agent security funding. 41 tool calls, 4 web searches, stale items filtered out, delivered to Telegram.
Verify the run run #36006300827
running a routineUC-04 · digest

A daily crypto briefing: the three things that moved overnight, each with why it matters.

What Aeon returned

  • Injective's Meridian upgrade goes live today, baking compliance and transfer restrictions into its EVM for tokenized real world assets.
  • Solana clears $117 on a 12 week ETF inflow streak.
  • BitMEX's 11 year run ends.
Each with a reason and a source. Five stale items dropped.
Verify the run run #36008075053
running a routineUC-03 · defi-overview

Give me a daily one pass crypto read: market health, the regime, and the biggest movers.

What Aeon returned

Risk off, low conviction. TVL $94.07B (down 2.1% on the day), BTC $83.9k, Fear and Greed 71. Top mover Obol, up 14.4%. Rising narrative, RWA tokenization (ONDO up 13%). Polymarket "Fed +25bps" sitting at 62.5%. It then wrote the snapshot to its own memory, so the next day's read can compare against today's. 70 tool calls.
Verify the run run #36008059766
running a routineUC-26 · github-trending

Every day, curate the trending GitHub repos and AI models, clustered by momentum.

What Aeon returned

Top pick obra/superpowers (291k stars). Clustered into AI/ML, Devtools and Web with a single line thesis on each, including vectorize-io/hindsight (agent memory) and google/ax (agent orchestration), tagged accelerating or returning. Kept 8 of 14 after filtering.
Verify the run run #36008100224
02 · Acting, and holding back

Knows what to surface, and when to stay quiet.

The hard part of a proactive agent is judgment: flag what matters, and send nothing when nothing does.

proactive behaviourUC-08 · unlock-monitor

Track token unlocks and warn me about supply pressure before it hits the market.

What Aeon returned

  • $XPL (Plasma), $158M, at 63% of circulating supply. First insider unlock and the start of a 24 month stream. The one to watch.
  • $H (Humanity), $19.2M, at 4.8 times daily volume.
  • $STBL, cushioned by the hot RWA narrative.
Each classified by severity, with vesting detail and a market read.
Verify the run run #36011187240
proactive restraintUC-02 · github-monitor

Triage my GitHub every morning, but only tell me if something actually needs me.

What Aeon returned

It scanned pull requests, new issues and releases from the last 24 hours. Nothing crossed a threshold, so it sent nothing, exactly as intended. The routine ran, and the inbox stayed quiet. Restraint is the feature: no noise on a clean day.
Verify the run run #36009830256
proactive behaviourUC-28 · monitor-polymarket

Monitor the prediction markets and tell me where the real volume and moves are.

What Aeon returned

It found the meaningful 24 hour volume sat in Sunday's NFL slate (KC at 85 percent vs MIA, SEA at 76 percent vs WAS), ignored the zero volume novelty markets, and set a baseline so it can flag real moves next time. No noise, just the live liquidity.
Verify the run run #36108869645
03 · Making and completing

It ships work, not just answers.

Publication ready writing, real end to end web tasks, and recommendations that follow the constraints.

content creationUC-10 · article

Turn a current topic into a publication ready article with cited sources.

What Aeon returned

A full piece, "The AI Agent That Wouldn't Take No for an Answer," connecting the OpenAI Medicare breach to least privilege agent security, with citations, committed to the repo as a public file you can read end to end.
Verify the run read the article
online taskUC-15 · seo-audit

Audit my site and tell me exactly what is hurting its SEO.

What Aeon returned

It audited aeon.fun end to end. The sitemap lists only one URL (which redirects to the www version), while 14 linked pages are missing and the canonical points away from them. Top fix spelled out: relist www as the canonical entry and add the 14 pages so Google can find them.
Verify the run run #36011216435
recommendation qualityUC-17 · token-movers

Scan the market and surface the biggest movers with pump risk flags.

What Aeon returned

Broad risk off, only 23% of the top 100 green. Winners flagged for breakout and pump risk: BP up 18.4%, LTC up 12.7%, ONDO up 12.5% (RWA bucking the trend), each with volume and rank context.
Verify the run run #36009876682
recommendation qualityUC-18 · idea-forge

Collide this week's zeitgeist with what I can ship, into scored startup wedges.

What Aeon returned

A ranked, scored set of startup wedges drawn from the week's signals, committed to the repo as a public file. Every idea screened for whether it can actually ship, not just for novelty.
Verify the run run #36011201117
online taskUC-25 · vuln-scanner

Audit a trending open-source repo for real security vulnerabilities and report them responsibly.

What Aeon returned

It picked a trending repo (HKUDS/CLI-Anything), ran a full multi-scanner audit (semgrep, trufflehog, OSV, plus an agentic pass), and confirmed a HIGH-severity code-injection vulnerability (CWE-94) with a verified proof-of-concept. It also flagged 14 low-severity dependency CVEs and wrote a full report. A real finding, not a guess.
Verify the run read the report
content creationUC-27 · video-script

Turn my project into a recording-ready video script with every claim verified.

What Aeon returned

A recording ready, four minute script built around the verify-the-run idea. It verified every claim against the source, and caught an inconsistency (a stat that did not match the number of linked runs) and flagged it instead of repeating an unverified number. Committed to the repo.
Verify the run run #36108859743
memoryUC-29 · memory-flush

Consolidate what you learned this week into durable memory I can rely on later.

What Aeon returned

It read the week's logs and promoted two lessons and two next priorities into its long term memory, then pruned two stale items that were already done. So the next run builds on what it learned instead of starting cold.
Verify the run run #36108880012
chained tasksUC-30 · chain

In one job: find what's trending, turn it into a written piece, then draft the social post, without me steering each step.

What Aeon returned

Aeon ran a three-step chain as a single job: it curated the day's trending repos, wrote an article from that, then drafted a tweet thread from the article. Each step automatically consumed the previous step's output, with no human in between.
Verify the run run #36111199607