Cyril Tovena 🇪🇺

We've been working on a @github integration for @grafana @PyroscopeIO. This is going help everyone understand where ressource is being consumed in your function. When you get the flamegraph like below your next question is where in that function ? github.com/grafana/pyrosc…

@edwinarbus @cursor_ai Yeah I believe this. If frontier labs were not selling at lost that would be a big deal.

Matt Maher tested frontier models in Cursor v. other harnesses. Cursor boosted model performance by 11% on average: Gemini: 52% → 57% GPT-5.4: 82% → 88% Opus: 77% → 93% His benchmark measures how well models implement a 100-feature PRD. @cursor_ai consistently outperformed.

Nice touch @OpenAI

@ryancarson My favorite one is using it with Grafana Assistant CLI to find errors logs, performance issue or gap in o11y and fix them daily. A must for new projects that needs to scale. I get PR everyday that fixes issues in dev and prod.

Trying out some new Automations in Codex

@Dimillian Definitely, mine does ui work and send me screenshots in linear and GitHub PR. I also setup automation in codex app that creates next item or problem found in my deployment via Grafana Assistant CLI and it creates linear items in my backlog then I just pick what’s next

Ok, symphony is truly a "feel the AGI" moment. I have PokeSwift mostly automated in my linear, I can just create issues, put them in todo, and the orchestrator does the rest. Then I just feedback on githuba I get a fix commit. Truly insane. I can solely focus on testing.

Been building a better UI for OpenAI Symphony.

@0xAkvn @joeyfreeland @grafana @RhysSullivan Yes and you can still connect your agent to the Grafana assistant via the assistant cli, both camp ⛺️

@joeyfreeland @grafana @RhysSullivan Exactly, why would I want to spend time connecting my non-specialized agent to Grafana when I can “just use Grafana Assistant”? Both options are there though which should be the case.

in app ai assistants mostly seem like a dead end to me i don't want to use your assistant, i just want your service to connect to my agent easily with all my other integrations apps are also incentivized to not spend on tokens making their assistance worse

@thsottiaux Review: it’s great but most of the time I need to ask about more details

Speed and front-end capabilities: we heard that loudly, and things are improving quickly, to the point where I think we are now highly competitive on speed, if not best-in-class. What else should we improve for Codex?

@sankeydadev @matryer Which one are you insterested in ?

@matryer @Kuqd It's refreshing to see a focus on UX, but what about handling edge cases? That’s where many tools fall short. Would love to hear how you tackled that in Grafana Assistant!

We made a video about how we made Grafana Assistant. Typescript frontend, #golang backend, some cool context/prompt tricks, and an obsession with the UX. It’s me, @Kuqd and Sven Grossmann. youtu.be/TGWVqBAeIzI

@krzyzanowskim Does it work with cursor cli ?

This is what peak Agentic Development looks like whether you like it or not.

Planning on Cursor with Opus 4.6 is just MAGICAL. They don't have the best model, but they at least have the best harness.

@thsottiaux Build a plan in a worktree from local

What could we do better on Codex? App, model, strategy and features… what’s wrong in how we approach things that we should improve immediately?

@thsottiaux Skills support per repo

@azlandotgg @cursor_ai More like 100

@cursor_ai so basically: blink and you owe $20

Opus 4.6 (fast mode) is now available in Cursor! It's 2.5x as fast in research preview.

@OpenAI We need a shortcut in the codex app to change reasoning may be cmd + t ? Also while on a worktree not always seeing the button to offload to local.

A2A is fantastic why nobody talks about it.

We recorded this long ago, I watched it again and it still so much relevant. A lot of good advice for building strong agent TLDR; KISS, keep it simple, stupid

No Codex 5.3 API. You’re forced into the Codex app. The app is so good you forget you were forced. Cursor sweating. Token costs suspiciously cheap. Compaction is insane. It codes like a senior engineer on espresso. OpenAI really said: “fine, I’ll do it myself” and actually did😅

@ericzakariasson @gregorian909 @cursor_ai Let me also build locally from a fresh context ;) I already can but manually

@gregorian909 @cursor_ai try this!

It'd be really helpful to create a plan locally and then hand it off to the cursor cloud agent @ericzakariasson @cursor_ai