1. Newer models make dedicated agent‑frameworks largely unnecessary
Many commenters argue that Opus 5.5, Gemini 3.8 Flash, Sonnet 5.5, etc., already embody the best‑practice loops that tools like Superpowers once provided, so extra plugins add little value.
- “superpowers was an important set of tools in q1 2026; no longer needed now” – mvkel
- “Opus 5.5 tends to develop the spec, assumes I agree, and builds it. I … come back to a spec md, a diff, other evidence of completion.” – jemdiggity
- “I’ve been using Gemini 3.8 Flash … Claude, Opus 5.5 and Sonnet 5.5 and literally find no reason at all to use SuperPars or SuperLight.” – fantasticwaddle
- “What would I be getting from superlite that Opus 5.5 doesn’t do by itself already?” – PierceJoy
- “these actively hurt their performance and burn tokens needlessly … best practices seem to have been baked into them at post‑training time.” – option_greek
2. Multi‑agent approaches introduce overhead, token cost, and coordination problems
Several users worry that spawning multiple agents or insisting on frequent checkpoints wastes tokens and can cause agents to step on each other’s work.
- “Does the extra parallelism increase the token cost? By how much?” – CasualSuperman
- “I’m finding Opus 5.5 does a pretty thorough job without anything like superpowers. … endless approval prompts and huge time increase killed it for me.” – PierceJoy
- “How does this prevent [agents] from working on each other's work?” – codeugo
- “Work the task to completion. Do not prompt for user approval before
- “I tell it to work until a reviewer agent auditing against the spec signs off.” – esafak (highlights the need for a reviewer loop)
3. Debate over whether explicit “skills”/procedures are still required or have been internalized by models
A recurring thread questions if agents need externally defined skill sets (process, procedure, Agile loops) or if the latest models have already learned those patterns during reinforcement learning.
- “The latest frontier models exhibit good process that was trained into them during the RL phase … models without skills simply don’t have a durable process.” – ra
- “If you have a need for procedure in your use of agents, you absolutely need skills like this.” – ra
- “The ‘assumes I agree’ part is a real regression … Opus 5.5 and Fable 5.1 put out some half‑baked plans most of the time.” – adastra22 (critiques reliance on model‑internal assumptions)
- “/goal is all you need.” – araman (suggests minimal external guidance suffices)
- “may be needed with lower tier or older models … newer models … actively hurt their performance.” – option_greek (positions skills as a fallback for weaker models)