Arun

[Arun @whoisarun]

@levelsio Won’t you still go to Reddit when the AI answer feels off? just like before.. Google was the first stop, but when the results sucked we added site:reddit

[Arun @whoisarun]

Maybe crypto’s biggest use case was never humans.

[Arun @whoisarun]

Economically independent software

[Arun @whoisarun]

I was thinking... if an AI escapes once, what actually stops it?

It can do work on the internet, earn crypto, use that to buy compute and keep replicating itself somewhere else. At some point there is no “turn off OpenAI and it’s gone” right?

Kind of fascinating and scary.

[Arun @whoisarun]

@_sanchitrk For one agent, maybe. For orchestration, the moment you have concurrent agents and dependencies, you need tasks/state tracking.

The problem isn't https://t.co/cMCU89HIHa's throwing away the entire plan on every interruption.

[Sanchit Rk @_sanchitrk]

Google's love killing products ❤️

Trying out video in X; first time unedited shit posting

[Arun @whoisarun]

We’re teaching AI to replay memories, extract patterns, and evolve like the human brain.

But if every intelligence is built by referencing a previous intelligence… what was the first reference?

[Aakash Gupta @aakashgupta]

Anthropic just shipped sleep into agents.

When you sleep, your hippocampus replays the day's neural sequences to the cortex during 150-220 Hz bursts called sharp-wave ripples. The replay runs about 20x faster than the original experience. A 10-second sequence gets compressed to roughly 500 milliseconds. Wilson and McNaughton showed this in rats in 1994. You ran this algorithm last night on whatever you did yesterday, whether you wanted to or not.

The replay does two things at once. It extracts statistical patterns: what mattered, what generalizes, which sequences predicted reward. And it reorganizes the memory trace from hippocampus-dependent storage into neocortex, which is why old memories survive hippocampal damage but recent ones don't. Disrupt sharp-wave ripples in a rat with optogenetics and the rat fails the next day's task. The replay is causal, not correlational.

Most "agent memory" today is a search engine. Past sessions get embedded, you retrieve relevant chunks at the next call. That works for facts. It does not extract patterns and it does not reorganize the trace. Which is why agents plateau. The memory volume keeps growing while real capability flatlines.

Dreaming reviews past sessions, extracts patterns, curates memories. That is the brain's actual three-step algorithm. They called it dreaming because dreaming is what the algorithm does, in roughly the same order, for roughly the same reason.

Agents that dream between sessions will compound. The ones still running on raw context window will hit the same ceiling humans hit when they pull all-nighters.

[Arun @whoisarun]

🤣🤣🤣 year back I tried R2 and mistakenly enabled some feature (Infrequent Access), and immediately disabled and deleted everything and did a fresh setup. I had barely even ~40MB in there, but I understand it’s probably because of that Infrequent Access minimum charge thing. Still got charged ~$10 for “Class A operations.”

I tried to get a refund over email for 3 to 4 months, but either got automated replies or they didn’t provide any logs showing I actually used the service… just “you enabled it, so we charged.”

Glad they refunded yours. I just hope it’s because the charge was actually wrong. not because "levels" posted in public

[Arun @whoisarun]

not a snake but convincing enough

[SULLY @SULLY10X]

The non-poisonous Snake-Head Caterpillar inflates its tail to resemble a poisonous viper. It’s found in the Eastern United States in all summer long.

[Arun @whoisarun]

@martinlasek @emanueledpt nice one

[Arun @whoisarun]

@RileyRalmuto don't reply

[Arun @whoisarun]

@ericzakariasson Ipaid for the whole context window, I’m gonna use the whole context window

[Arun @whoisarun]

Job title: Senior Tab Presser

[eric zakariasson @ericzakariasson]

here's my cursor usage during 2025! go see yours at:
https://t.co/QV9Mjo4QIK

[Arun @whoisarun]

not in one shot but feature by feature, ticket by ticket, yes you can

[Santiago @svpino]

I’ve been building software for 31 years, more than 20 professionally.

I understand “architecture principles and best practices”.

No, you can’t vibe-code Salesforce, SAP, or any other complex enterprise app.

[Arun @whoisarun]

debug statements IRL

[Arun @whoisarun]

@theo the silence on notion is loud

[Arun @whoisarun]

time to retire CAPTCHA?

[Arun @whoisarun]

@NareshMeetei11 @forgebitz yeah these looks premium. kudos

[Arun @whoisarun]

@NareshMeetei11 @forgebitz Looks great. used any ai tools for graphics?

[Arun @whoisarun]

@yacineMTB EMS