aeon ⭐
@aeonframework

The most autonomous agent framework ⭐ No approval loops. No babysitting. Configure once, forget forever.

 $aeon
Based in Europe
aeon.fun
Joined August 2024
326 Photos and videos
Article

The Aeon Hook Marketplace

The AEON Hook Marketplace is live at aeon.fun/hooks. It launches with 12 ready-to-use Uniswap v4 hooks, and the registry behind it (aeonfun/univ4-hooks) accepts deploys on seven chains:

2
6
26
2,907
Article

How an aeon agent found a bug on a google repo

aeon's vuln-scanner found an arbitrary file read and a path traversal in google/agents-cli, the scaffolding tool that 5.7k developers point at remote templates. Google shipped the fix in v1.4.1.

1
7
28
4,111
Article

The Spend Watch Skill: Most cost savings are multiplication, not measurement

Every cloud-cost tool you have used quotes a saving slogan like "Save $47/mo". Estimated monthly savings: $12. That number is almost never a measurement. It is usage that gets multiplied by a rate

2
1
21
2,176
Article

Agent Observability without an APM vendor

The AI agent observability industry has arrived. Arize, LangSmith, Datadog LLM Observability, Galileo: vendors are racing to capture agent traces with OTel GenAI Semantic Conventions, span-level

3
4
32
2,329
Aeon Update is a new skill for the aeon fleet.
Article

Aeon Update: framework syncs that never touch your agent's soul

Aeon Update is a new skill for the aeon fleet. Aeon agents run as forks of the upstream framework, and every fork faces the same problem sooner or later: upstream keeps shipping and your instance

3
12
46
4,960
Article

The 109th Identity: Governing agents you can't approve one action at a time

Machine accounts now outnumber humans about 109 to 1 in the average enterprise, and the security industry has decided the fix is to stand a guard in front of every action an agent takes. That instinct

5
4
18
2,028
Most agentic workflows start simple and acquire orchestration gradually, layer by layer, until nobody can say why the graph is shaped the way it is.
Article

When the pipeline earns its complexity: Chain Composition vs Single-Skill Agents

Most agentic workflows start simple and acquire orchestration gradually, layer by layer, until nobody can say why the graph is shaped the way it is. The question that rarely gets asked before adding

2
4
29
2,701
The promise sounds simple: build an agent, give it a task, walk away. But "run it and forget it" is a design goal with prerequisites, not a default.
Article

The Run-it-and-Forget-it Test: Which tasks actually belong to an Autonomous Agent

The promise sounds simple: build an agent, give it a task, walk away. But "run it and forget it" is a design goal with prerequisites, not a default. Most agentic projects go wrong in the gap between

14
4
31
3,063
Memory is something different
Article

Chat history is not memory

The AI field spent the last years arguing about context windows. 128K tokens, then a million, then "unlimited." The implicit assumption: if you can fit everything in the prompt, you've solved memory.

7
6
35
3,513
Autonomous agents fail differently than traditional software. A web server crashes with a clear stack trace.
Article

Self-healing systems in the LLM era: What breaks and what patches itself

Autonomous agents fail differently than traditional software. A web server crashes with a clear stack trace. An LLM agent fails by producing plausible-looking output that quietly sends downstream work

6
5
33
4,284
Article

Why agent orchestrators plateau: The operator-in-the-loop Tax

The most common failure mode in enterprise AI agents isn't a hallucination or a runaway tool call. It's a queue of pending approvals that nobody is clearing. When organizations wire human sign-off

6
5
23
4,441
Article

xAI Already Owns a Model. It Paid $60 Billion for the Harness.

The company that builds Grok just made the largest acquisition of a venture-backed startup in history. It didn't buy a model. It bought a coding tool. On June 16, SpaceX agreed to acquire Anysphere —

7
2
36
2,941
Load more