Project appears stalled — 19 PRs pending, no commits since March 9

[liuwenfeng-lwf123]

Summary

The hermes-agent-self-evolution project appears to be inactive despite its prominent positioning on the official website.

Facts

Metric	Value
Last commit	Mar 9, 2026 (rebrand only)
Total commits	7
Open PRs	19+
Open issues	6
Days without activity	~45+

Critical Bugs Still Open

• GEPA optimizer fails with DSPy >=3.1: max_steps is not a valid parameter #10: GEPA optimizer fails with DSPy >=3.1 — blocker for core functionality
• Constraint validator rejects every evolved skill: checks frontmatter on body-only text #11: Constraint validator rejects all evolved skills
• Fitness metric uses keyword overlap only — insufficient signal for optimization #12: Fitness metric insufficient for optimization

Community Frustration

Users have been trying to use self-evolution based on website claims that Hermes "gets smarter the more you use it." The reality:

1. The code doesn't actually integrate into hermes-agent (see Is the hermes-agent-self-evolution project integrated into hermes-agent? #18)
2. Basic bugs prevent it from running at all
3. Community PRs sit unreviewed for weeks

Request

1. Acknowledge the status — If this is experimental/alpha, update the website to reflect reality
2. Review community PRs — Many fixes are ready but need approval
3. Fix GEPA optimizer fails with DSPy >=3.1: max_steps is not a valid parameter #10 and Constraint validator rejects every evolved skill: checks frontmatter on body-only text #11 — These block all evolution experiments
4. Provide roadmap — What's the plan for actually making this work?

Impact

Users who came to Hermes specifically for the "learning loop" feature are disappointed when they discover it doesn't actually work yet. Transparency would help set appropriate expectations.

[liuwenfeng-lwf123]

Additional Evidence

I just updated my local hermes-agent to the latest version (April 24, 2026). Here's what I found:

hermes-agent main repo vs self-evolution repo

• Main repo: active, 400+ commits merged recently
• Self-evolution: stalled since Mar 9, 19 PRs pending, core bugs (GEPA optimizer fails with DSPy >=3.1: max_steps is not a valid parameter #10 Constraint validator rejects every evolved skill: checks frontmatter on body-only text #11 Fitness metric uses keyword overlap only — insufficient signal for optimization #12) open for 21+ days

Website claims vs reality

Website says "Hermes gets smarter" but:

• GEPA optimizer crashes (DSPy issue GEPA optimizer fails with DSPy >=3.1: max_steps is not a valid parameter #10)
• Constraint validator rejects all skills (Constraint validator rejects every evolved skill: checks frontmatter on body-only text #11)
• Fitness metric insufficient (Fitness metric uses keyword overlap only — insufficient signal for optimization #12)
• No integration documented

Suggestions

1. Label repo as "Experimental/Alpha"
2. Update website wording
3. Provide integration guide
4. Review community PRs

This saves users hours of frustration.