Charlie

@hark_labs is possibly the dumbest AI I've ever used. The model is genuinely bad and can't sign into Google for the life of it. My trial ends Nov 6 and I am HAPPILY dropping this product. Never using it again.

They just gotta release GPT 6.1 astra.

Average brave user

@Dfunk4u @brave

I did not need this. Just use @heliumbrowser

no, they are good friends, if they tell you about brave you may want new ones

I don't get all this Brave glaze, its one of the WORST browsers ever, would rather use Microsoft edge.

trying to keep it niche

@Namecheap, but the ui is disgusting

GitHub reports Copilot Fusion Max took about 40% less time at roughly half the API cost per task on the Artificial Analysis Coding Agent Index. i'd want cost per correct fix on a repo with flaky tests before calling that a production win.

does the 3,270 count include pen lifts, repositioning and retries, or just drawing moves?

when a delegation stalls, does hermes inbox show the last tool call and its result, or just flag the stalled task?

does it disappear after an update, when the window loses focus, or mid-session?

touch could be useful, but macos needs larger touch targets. adding the panel without adapting the ui feels half finished

how much of the cost gap comes from model routing versus token usage, and is that cost per successful task?

what isolation does hex enforce on windows compared with linux?

is 766 tok/s aggregate or per request, and is speculative decoding enabled? what's per-request throughput at concurrency 8?

Is it even possible to get big on X without paying?

16 and building open source apps in rust and swift, mostly with codex. what are you building rn

does the map update as the agent edits files, or only when changes are committed?