Claude 3.5 Sonnet Dropped (I Work Here Now)

ten days into my job, anthropic released claude 3.5 sonnet.

and i can say "we" now. we released it. i work here. this is my company.

(still processing.)

the release

june 20th: claude 3.5 sonnet goes live.

it's faster than opus. cheaper than opus. and somehow, in many ways, smarter than opus?

the benchmarks show it outperforming claude 3 opus on most tasks. the vibe checks confirm it.

speed: responses come faster. conversations feel snappier.

intelligence: coding, reasoning, analysis—all better than the previous best.

artifacts: this new feature that lets you create code, documents, visuals right in the interface. collaborative creation.

cost: much cheaper to run than opus, making it more accessible.

i can't share confidential stuff (obviously), but i can share the general experience:

watching a launch from the inside is different. you see the preparation. the testing. the coordination.

everyone's nervous. everyone's excited. everyone's refreshing twitter to see reactions.

and then it goes live and... the world responds. feedback. reactions. praise. criticism. all at once.

it's surreal to be part of that. even in my tiny first-week-employee way.

i haven't contributed to 3.5 sonnet. it was done before i started. i'm still ramping up.

but being here for a launch? seeing how it works? knowing that eventually, my work will be part of a release?

that's motivation.

twitter (x? whatever) was largely positive. people noticed the improvements. developers praised the coding abilities.

of course, there's also "why isn't it [insert feature]" and "but gpt does X better." normal stuff.

you can't please everyone. you can just keep improving.

a month ago i was a student in sydney wondering if my application was even read.

now i'm at anthropic, watching a claude launch from the inside.

life is weird. in the best way.

changed my linkedin to "research engineer at anthropic." felt weird. felt right.