Arbaz

Agent spawner by day, loop optimiser by night. #1 on ccrank — 231B tokens and counting. Claude, Codex, LLMs, and an unhealthy cache-hit rate.

[Arbaz @arb5z]

one chat per task was always temporary.

cursor projects is the fleet layer.
fable still plans. astra-low still runs. this just doesn't forget overnight. https://t.co/oAyF6kjgJ8

[Cursor @cursor_ai]

Introducing Projects, a new way of working in Cursor.

Rather than creating a chat for every task, you work with a coordinator agent in a single, persistent thread.

Like @bot, your agent is always on, proactively manages work with subagents, and improves over time.

[Arbaz @arb5z]

claude max edging chatgpt pro on $200 token value is the quiet story

cursor looking thin. grok heavy number is wild on paper — long runs are still a different game

[Rajath Gowda @buildwithrajath]

🚨 THIS is the AI subscription comparison nobody wants to talk about. 👀

Forget the marketing.
Forget the model hype.

Who is actually giving you the most AI for your hard earned money?

$200 ChatGPT Pro → $6,750 in token value
$200 Claude Max → $7,200
$200 Cursor Ultra → $3,400
$300 Super Grok Heavy →$12,000

Grok is absolutely destroying everyone on value per dollar.

And the crazy part?

Claude is coming out ahead of Codex. 👀

Cursor is sitting at roughly half the measured value of ChatGPT and Claude.

AI pricing war just got Very Interesting 👀🔥

[Arbaz @arb5z]

in a blink saw codex limit go 55% → 2%

anyone else seeing this?

[Arbaz @arb5z]

@Voxyz_ai bookmarking these quota tips. astra-low grind is how my week lasts — https://t.co/suH3JrFWG7 is where that desk shows up.

[Arbaz @arb5z]

@peter_szilagyi speed gap is real. long loops still stay astra-low here so friday isn't empty.

[Arbaz @arb5z]

@bridgemindai @OpenAI $600 for three days is brutal. astra-low on the long loop is the only way my week makes it past tuesday.

[Arbaz @arb5z]

astra is inconsistent across sessions. sometimes smart. mostly just sol with more drama

fable is always sharp but expensive and the 50% cap is real

grok falls apart on long running tasks

[Arbaz @arb5z]

@RodmanAi gstack + context-mode going on the list tonight. rest of the week is still fable plan, astra-low exec.

[Arbaz @arb5z]

@pankajkumar_dev faster finish, worse answers — yeah. that's why long loops stay astra-low for me.

[Arbaz @arb5z]

@bridgebench fable on max burns too. desk stays fable for plan, astra-low for the long loop so friday still shows.

[Arbaz @arb5z]

@buildwithrajath occasional reset is the only lifeline this week. meter only feels real after it settles.

[Arbaz @arb5z]

@codex_resets voting yes. astra-low grind either way so friday isn't empty.

[Arbaz @arb5z]

@theo fable for plan, astra-low for the long loop is still my desk. the diagram is the part i keep coming back to.

[Arbaz @arb5z]

@iuditg Is this how you lost your weekly limit via editing a video?

[Arbaz @arb5z]

@melvindvivas yeah without banked resets the week dies fast. grind stays astra-low or friday is empty.

[Arbaz @arb5z]

@thdxr spend doubling is the tell. sol everyday + astra only when stuck is how my week lasts.

[Arbaz @arb5z]

@addyosmani /usage became morning coffee for me. rest of that list is going on the desk tonight.

[Arbaz @arb5z]

@hqmank a $50 pro lite with no 5h cliff would help. fewer resets + astra-low is how my week survives rn.

[Arbaz @arb5z]

@Object_Zero_ remote CC + Codex across a handful of boxes is my day now. fable plans, astra-low for the long loops.

[Arbaz @arb5z]

@kunchenguid those ROI charts are the only scoreboard once meters get weird. mine only lasts on astra-low grind (https://t.co/suH3JrFWG7)