Getting insights from thousands of offline customer conversations used to be next to impossible - slow, manual, and expensive. With Rilla Intelligence, it’s now instant.

he used to hate python but @astral_sh and @charliermarsh are just too good

Easily the hardest project I’ve ever worked on, but also the most rewarding. Excited to see where we can take it next.

For online commerce, this level of insight is table stakes. But 90% of commerce still happens offline - and until now, no one could measure it.

Customers are already using it to figure out – why they’re losing deals to competitors – which brands people like most – what top reps are doing differently and much more

The documentation recommends with_options to set timeout or max_retries per request and doesn’t warn about this at all.

So this week I learned that client.with_options in the @OpenAI python library actually creates a completely new client. This caused us a ton of performance issues because we were sending 1000s of parallel requests and now each of those had to establish a new HTTP connection.

https://t.co/Ux4WebvH7h

Absolutely love the “toilet paper roll test” from @scarletinked’s latest article. If you are at someone else’s house and notice the toilet paper roll is backwards - would you fix it? The very best employees are the ones who would fix the roll.

Love this @datadoghq feature. Calling bs on the @OpenAI status page.

T-3 days until the Rilla Invitational. Time to fully transform this warehouse.

This is what launch prep looks like behind the scenes 👀 Demo squad assembling the fleet for Monday.

“This will just be a small prompt tweak.” 4 hours later: one bug fixed, two new ones created.

Trying a pattern right now where the LLM only has one tool to execute code (using @e2b). All other tools are helper functions or API calls inside the code environment. Works very well so far because the tools compose much better.