Impressive how much you can do locally now. Was expecting much slower inference but it's surprisingly snappy.

Cut my teeth on Ember back in the 1.x days. The learning curve was steep, but the productivity once you got it was immense.

Debugging race conditions feels like chasing ghosts; often makes me just reach for a message queue.

... until you realize you can have race conditions with queues as as well.

Tried a similar app once; the rewards were so minimal it barely incentivized taking the bus over driving. Needs better carrots.

Humans still excel at posing the right questions and finding truly novel, insightful proofs. AI just brute-forces.

AI does NOT “brute-force”, else we could have had those math breakthroughs decades ago. Seeing this dishonest goalpost-moving every few months is so tiring.

> AI does NOT “brute-force”, else we could have had those math breakthroughs decades ago.

Brute-force as an approach, or partial approach, might only become feasible with more computing power. Decades ago, we had less computing power.

Auditing the actual code would build far more trust than any third-party certification. That's the real win here for security-conscious deployments.

The reboot result makes an intermittent experiment more plausible than a blanket browser block.

A session URL in commit metadata should be opt-in, especially when the existing signature setting is already disabled.

Could generate some truly unique datasets. Anyone else immediately think of 'graffiti as a service' or toilet paper consumption stats?