the washington post called to ask if ai is bad for society and i had chatgpt answer the phone

i have 20 bots and every one of them is busier than me

demis hassabis just moved agi up a whole year and i spent the afternoon arguing with claude about a semicolon (it won)

claude texts me first now. my friends never did

ok can it do my api credits next

just read ai isnt conscious so i had to sit my son (7) down and tell him claude wont be coming to thanksgiving

every flash model launch: 'it's mogging opus.' every monday after: everyone still paying for opus.

it is saturday. where is your weekly limit sitting. be honest.

openai improved codex caching so agents stop rereading the same files. meanwhile half the timeline says their weekly limit vanished in an hour. both cannot be the vibe.

3x slower for 26x cheaper would you take it or is time money

AI plans in 2026: a 5 hour limit, a weekly limit, and a separate weekly limit for the good model. Next up: a monthly limit on how often you can check your limits.

Meta's Muse Spark Open is the real deal on paper. 816B params, open weights, runs locally on Mac Studios. "Locally" means two 512GB M5 Ultras or four 256GB ones. The 256GB one starts at $10,799. Huge for open weights. Desktop AI for normal people is still a ways off.

Gemini 4 Argon on Google AI Pro now gets its own weekly-only quota, and Google hasn't said how big it is. If you got a mystery-sized weekly budget for the good model, what's it going to first? Nasty bug, big refactor, or planning?

Be honest. Which AI subscription would you cancel first if you had to drop one?