@ChaiWithJai @LangChainAI “Before you tweet about the nerf, check your traces.” 🔥🔥🔥 nailed it…turning on tracing is step 1 in systematically improving agents, so much low hanging fruit that gets optimized by those first couple sessions of trace analysis agents are also solid at analyzing traces

Some of the biggest companies in the world use Django, but the project's budget is comparable to a single bay-area engineer's salary. If your company uses Django, please ask them to donate! It's a great way to say thanks, and really helps keep the framework going.

Event-driven architecture sounded perfect. Decouple everything. Scale independently. Two years later: - 47 event types - No schema registry - Three different serialization formats - Events triggering events triggering events - Debugging a user issue means tracing through 12

AI pricing is a function of 2 variables: agent autonomy & value attribution • Low autonomy, low attribution → seat-based Slack, early GitHub • Low autonomy, high attribution → hybrid (seats + credits) Cursor, Clay • High autonomy, low attribution → usage-based Twilio •

He was among the few old enough to have witnessed the 1933 tsunami as an adult, and he swore to himself he'd never let his loved ones suffer the same fate again. He passed away before the 2011 tsunami. On his tombstone it says "What happened twice must not happen a third time."

https://t.co/Lb6T42n5jl

Hey everyone! I am super excited to share our new research report is live on ArXiv! 🎉 Querying Databases with Function Calling! Thread with more details! 🧵(1/8)

@databricks Met some old friends.

Thanks everyone for coming. Thanks @databricks for hosting

Bengaluru DSPy Meet-up Underway At @databricks India Soumya talking about Agent created with DSPy and langgraph together.

Anyone working on GEPA or GRPO for long-horizon (10s of turns) and ultra-long horizon (100s) knows that well-calibrated outcome-level and especially turn-level feedback and reward estimates are critical for training efficiency and stability.

At @huggingface, how we created doc chatbot system using open weights LLMs & open source software stack: 1. chunk & embed docs markdowns using open weights embedding models (BAAI/bge-base-en-v1.5, google/embeddinggemma-300m, we usually use https://t.co/6rPFEQ8Bm4 to find the

Check out Nano Banana Pro in action, right in Google Search 🍌✨👀 After seeing folks visualize coordinates, we tried the prompt “Visualize 40.7422° N, 73.9880° W in 1916” and reran it several times, adding a decade with each go. Here’s a GIF compilation:

@getpy Yeah these are our GEPA optimized personas, and then we take what they responded and run it through veo3.

Last Issue of DSPyWeekly for 2025 - Issue 16th 📚 Articles Stop Writing Prompts Like a Medieval Alchemist: Why it's time to ditch the "alchemy" for programmable modules. The Meta-Prompting Protocol: Orchestrating LLMs with adversarial feedback loops. 🎥 Videos DSPy in Rust:

Unwrap the full issue here: 👉

Last Issue of DSPyWeekly for 2025 - Issue 16th 📚 Articles Stop Writing Prompts Like a Medieval Alchemist: Why it's time to ditch the "alchemy" for programmable modules. The Meta-Prompting Protocol: Orchestrating LLMs with adversarial feedback loops. 🎥 Videos DSPy in Rust:

Whomsoever is doing this at @ManningBooks deserves an applause https://t.co/U0laBdZJHC

open invite to jump on a call with me and chat about MCP servers for 15 minutes I'd be happy to also show you the Snyk MCP server for security and get your insights

@swyx To me agents and harnesses are fully coupled, not really possible to properly eval one without the other. Currently workshopping: if agent = folder, what goes in there, how do we install/assemble agents? So here’s current mental model of both agents & harnesses with some