@MarioNawfal Uncapped AI spend is an increasing problem. You can cap per-key/org platform side (would have prevented this), or use proxy layer enforcement with LiteLLM or equiv. As a consultant I couldn't find a good fit, so I built one for Cursor as a MCP layer: https://t.co/KswYYGFJo6
👨🏭Iterating Towards LLM Reliability with Evaluation Driven Development
If you've interacted with the LangChain GitHub repo, you may have noticed a helpful GitHub bot 🤖called Dosu
You may have also noticed that it's been improving over time! In this blog @devstein64 dives into how they've used evaluation driven development to help them ship with confidence
https://t.co/KCYkDWShWn
Spending a couple weeks working with @devstein64 & team at @dosu_ai on code retrieval, so I thought I'd give them a shoutout!
Dosu basically does what I've been doing since RAGatouille came out: scour GitHub issues and desperately try to figure out what's causing them. Link below if you want to check them out
@ashley_in_tech As someone without a formal degree I can guarantee the search is hard. However as many other have said; networking and building up the personal project portfolio is key. Networking with smaller startups may be beneficial as well as they typically have less $$ + access to talent.
@tdave67@laurieontech@geo_walters I get that from a software standpoint it’s crucial. However using it in terms of saying “hey let’s have a meeting at 3pm PST” over a Slack message is likely going to be understood regardless of if the current pacific timezone is really PDT.
We tried our best to create a defensible space in front of our condos on Daggett Summit before evacuating earlier today. Wishing our firefighters the best as they fight this.
#CaldorFire
@chamath I’m all for electric, but you could likely drive across Norway in 1-2 charges of your vehicle. The same cannot be said for places like the United States. I think battery technology needs to get even better before we can see numbers like these.
@NTTL_LTTN I use containers to reduce dependency headaches for other developers trying to work on the codebase. I don’t use them everywhere, but they have their place.
They also are great for scaling horizontally as long as the underlying code/software & architecture is setup for it.
@NTTL_LTTN We had a conflict in our GraphQL schema on production only. Basically a untested endpoint slipped through that was out of sync with the backend schema. Rolled back the commit and setup some testing for the endpoint.
Lesson learned to write solid unit tests for all API endpoints.
I’m very excited to finally announce I’ve accepted an offer to join Tellus as their Full-Stack Software Engineer! 💻🍾
I’ve had a blast contracting full-time with them the past few months and can’t wait to see what this year brings!
https://t.co/3Dg6fLrXUd