benjbrooks

I'm happy with OpenRouter. They're good at routing requests & spend tracking/management. You should just expect to directly own your relationship with the downstream model provider.

Source: Enterprise customer doing $XXM annual run rate of inference spend on their platform

disclosure: i'm an eng on this team

sales is a really interesting breeding ground for the first wave of vertical agents:

- success metrics are well-defined (did we get an email response? did we book a meeting? did we move a deal down the funnel?)

- enterprise sales is about getting a reasonably well-defined process working at scale. sales reps have to deeply understand enablement materials (what problems does my employer solve? what other companies have those problems?), qualify prospects relentlessly, and talk to lots of potential buyers.

- it's knowledge work (writing emails, preparing documents, communicating with counterparts over video call, etc.) that is mostly about aggregating information at scale. the really good sellers take rigorous notes and ask thoughtful questions.

Assuming they have the right information in context (hard work but not complex), LLMs are quite good at this work. And we think we're taking a pretty unique position in that we're not trying to replace top-of-funnel email drafting (this is somewhat trivial & also not that useful), we're making it significantly easier for sellers to find & utilize the information that enables them to do their jobs effectively.

I summited Mt Whitney last fall, a trail that 1) is notoriously hard to get a permit for and 2) requires all waste to be packed out via Wag Bags.

I was surprised to see a dozen or so wag bags tossed to the side of the trail over the course of my trip. You’d think that visitors would either poop on the ground with no regard for others or pack out their waste, not take all the effort to bring the bag but leave the remnants. It really left a sour taste in my mouth (and smell in my nostrils).

o3 has yet to fail me on complex, multi-table queries. Not a fan of BigQuery’s Gemini integration.

would love to use this but def need soc2 to justify it to our security folks

I was trying my best to finish quickly, initially in hopes of making the NYC Marathon. The "best" way to do the trip is probably to take everything in and have a few zero days, potentially bake in some time for wandering. I think part of me was worried that if I stopped, I wouldn't have the gumption to start up again.

Sounds like you had a great experience riding too! Did you go across the U.S.?

I think I'll pause the bike tour travel for the foreseeable future. I think I like running better :)

check out the map! https://map.brooks.team

i was a little worried about bears for the night or two i was in bear country but my fear of cars and weather was far more top of mind

i didn't look into different tires. my hypothesis is that most of my flats can be attributed to all the weight being on the back tire.

favorite part was jumping into extended conversations with strangers. from a scenery perspective, coming down into Lake Tahoe from Eldorado was just absolutely stunning. same when I went past Bryce Canyon.

We’ve been running our company’s friends and family newsletter on buttondown for more than a year and love it! Congrats on the domain purchase.

You may consider starting with what you consider to be a reasonable price and doubling the price for every subsequent customer until they start to churn during pricing discussions. This is particularly useful if you’re doing B2B sales, less so if you’re putting the price on a landing page.

Update:

Thanks for your recommendations everyone! We decided to go the route of measuring successful hits on the endpoint associated with our docs (inside our monitoring service). That's the default health check associated with our load balancer, so it gets hit periodically (built-in cron job). We just added a signoz alert that is triggered if the sum of those successful calls over the past X seconds falls below a threshold.

on second read, this is something that is set up on the service-side. I.e. if our service goes down, we would also stop sending data to Signoz Cloud. We wouldn't be able to run a status check, let alone send it to signoz cloud.

ah, this is the answer I was hoping for. thank you!

thanks! this has been recommended by others. I'm not feeling excited about migrating to another monitoring service but it might be the right move.

Disclaimer: I’m one of the founders

We’re working on text/image prompted vision models at DirectAI (https://directai.io). We help clients do detection of bespoke objects by defining them with a single phrase or prompt image. I’d be happy to talk about helping you build out a system like this without having to collect large amounts of data, annotate it, and then train a bunch of custom models.

helpful context, just sent for discussion in our team slack

this is sick!