Emmanuel

[David Cramer @zeeg]

I am here once again to say “put up or shut up”

One million to open source this year, as we continue to give back to the ecosystem via the pledge.

More than half has gone directly to your favorite open source developers.

https://t.co/kSHKM2BxJL

[Y_B @Yann_Bilien]

https://t.co/6jDVc7yV1H

[Emmanuel @Littleyoda110]

@morganlinton @daedalium Agent harness ≠ solution.

It’s complexity layered on top of non-determinism.

More loops, more tools, more memory
→ still no guarantee.

Real question:
who makes the decision and can you trust it every time?

[Emmanuel @Littleyoda110]

@randomtryidk Super sharp breakdown. “The harness is the product” really nails why Claude Code feels different. This is where the space is going: not just smarter models, but better runtime, trust, and UX.

[Thomas Molinier @randomtryidk]

https://t.co/3dfUSQavrl

[Y_B @Yann_Bilien]

https://t.co/8vo4oFSiPW

[Y_B @Yann_Bilien]

@Voxyz_ai @garrytan Yep true, what’s really powerful is to keep in memory those skills output to enforce them again and again without me asking again
Sharing publicly what I use

[Emmanuel @Littleyoda110]

@random_walker @steverab This paper nails the core problem. Agents get more capable but not more reliable. The missing piece is deterministic enforcement between what an agent proposes and what actually executes.

[Emmanuel @Littleyoda110]

@Clara_Gold Context is everything !! Always

[Rippletide @RippletideCo]

Support agents: 80% → 95%+ tickets correctly resolved.
Generated UIs: 90% → 99.99% design system compliance.
Code agents: 25% → 75%+ PR acceptance rate.

Same models. Same prompts.

One difference: an enforcement layer that validates every decision at runtime.

[Y_B @Yann_Bilien]

@thsottiaux long term memory! The one that remembers my patterns and edits in plans/sessions/PRs and apply it directly to new ones without me intervening again. It’s currently our main blocker for having 4+ codex running in parallel. Even more when I use codex as an orchestrator for my VMs background agents. I think a graph memory is a path, happy to chat

[@jason @Jason]

Come on the pod @DarioAmodei

[Rippletide @RippletideCo]

https://t.co/xlLP1Ob8wD

[Emmanuel @Littleyoda110]

@RippletideCo 👏👏👏 Well done team

[Rippletide @RippletideCo]

AI agents fail quietly. That’s the scary part.
Nothing crashes. You just get a reasonable answer… that’s wrong.
No trace. No explanation.

We shipped Rippletide Eval CLI to change that.
Explainable. Repeatable. CLI-first.

Live on Product Hunt 👉 https://t.co/cmVfaE6R1w

[Emmanuel @Littleyoda110]

@daedalium Yeah, the experience will be very different….

[Emmanuel @Littleyoda110]

@Pascal_Laurent_ Pas besoin de partir si loin, rester en France suffit …

[Emmanuel @Littleyoda110]

@Paul_Theway Rester fort et tenir toutes ses positions ou alors sortir totalement du marché!
Le marché est compliqué, pas de place pour la demi mesure rendu à ce moment du cycle je pense.

[Emmanuel @Littleyoda110]

@CerfiaFR A ce rythme là qu’il arrête l’obligation d’aller a l’école jusqu’à 16 ans…