Dante Lopez

dev in manila

[Dante Lopez @dqlopez]

Cozy night with @opencode & @MiniMax_AI.

[Dante Lopez @dqlopez]

@capsraunak

[Dante Lopez @dqlopez]

@guenolekikabou @thsottiaux I unsubscribed already but still have access to it until this week. I settled for Cursor for now, at least I can grind on hours that I can.

[Dante Lopez @dqlopez]

@danellisona Briefly introduced to assembly then C

[Dante Lopez @dqlopez]

@Quiply @thsottiaux bad faith of what? He asks I just answered facts.

[Dante Lopez @dqlopez]

@VulcanBench Any plans on including Devin harness in your benchmarks?

[Dante Lopez @dqlopez]

@jpschroeder There's something in blue that makes me somehow daydream

[Dante Lopez @dqlopez]

@morganlinton Thanks Morgan!

[Dante Lopez @dqlopez]

@ericzakariasson @bot https://t.co/I8sCKwedoo

[Dante Lopez @dqlopez]

Is this a Grok Bot limitation for multi-repo GitHub watches (via routines)? Each bot run one at a time, not in parallel per repo.

I tested 8 repo watches with simultaneous PRs across different orgs. Across 20 runs, zero overlapped and each started the same second the last one finished. Chat still works during a watch, so this seems limited to isolated watch executions.

poteto poteto poteto you there?

[Dante Lopez @dqlopez]

@VulcanBench Muse Code is out of beta. Can you also do benchmark of Muse Spark on Pi vs Muse Code?

[Dante Lopez @dqlopez]

Meta's Muse Code subscription plan has been released. Honestly, this is cheap.

[Dante Lopez @dqlopez]

'The red X is real. That 8:28 PM run died in 52 seconds with resource_exhausted / provider overloaded. Grok Bot’s model was over capacity, not a bug in the watch itself."

anyone else experiencing Grok Bot downtime?

[Dante Lopez @dqlopez]

@kr0der I did couple of nights ago, bit drunk, I misspelled 😂

[Dante Lopez @dqlopez]

There seems to be a Grok Bot limitation for multi-repo GitHub watches, as each run one at a time not in parallel per repo.

I tested 8 repo watches with simultaneous PRs across different orgs. Across 20 runs, zero overlapped and each started the same second the last one finished. So a 4 minutes review on one repo delayed another repo's acknowledgement by four minutes. Chat still works during a watch, so this seems limited to isolated watch executions.

[Dante Lopez @dqlopez]

Is this a Grok Bot limitation for multi-repo GitHub watches (via routines)? Each bot run one at a time, not in parallel per repo.

I tested 8 repo watches with simultaneous PRs across different orgs. Across 20 runs, zero overlapped and each started the same second the last one finished. Chat still works during a watch, so this seems limited to isolated watch executions.

poteto poteto poteto you there?

[Dante Lopez @dqlopez]

Its been a while since I manually compact conversation. I just let the harness do its thing when to summarize. I still read the code so I can say there's little to none of degradation on my workflow. The only catch is that prior/early conversations tends to be missed so you have to re-iterate.

[Dante Lopez @dqlopez]

If you're in $20 plan, Codex is really not the perfect fit, especially if you are working on sides or a weekend builder. The 5-hour limit disrupts your plan for the day.