The Gemini 2.5 Flash 05-20 variant is now the stable model we plan to support long term for Flash, and based on developer feedback, we have simplified the pricing and introduced an even smaller variant optimized for cost. (4/N)
LLMS
-
World’s Best AI Model Lineup Released Today
By
–
Huge work by the team for many months to bring the worlds best model lineup to everyone, I could not be more excited about today (2/N)
-
Google Introduces Gemini 2.5 Model Family Updates
By
–
Introducing the Gemini 2.5 model family: – Gemini 2.5 Pro (Stable, no changes from 06-05)
– Gemini 2.5 Flash (Stable, updated pricing from 05-20)
– Gemini 2.5 Flash-Lite (Preview, small reasoning model) More info in -
Google Launches Gemini 2.5 Flash and Pro Models
By
–
It's a BIG day for Gemini 2.5 — 2.5 Flash and 2.5 Pro are now stable and generally available in AI Studio, Vertex AI, and the @GeminiApp — We're launching a preview of the new 2.5 Flash-Lite, our most cost-efficient and fastest 2.5 model yet More info on each model below
-
Meta Announces First Llama Startup Program Cohort Selection
By
–
The response to our first-ever Llama Startup Program was astounding, and after reviewing over 1,000 applications we’re thrilled to announce our first group. This eclectic group of early-stage startups are ready to push the boundaries of what’s possible with Llama and drive
-
Morningstar Builds AI Research Assistant Mo Using LangGraph
By
–
Investment research, accelerated. @MorningstarInc built Mo, their AI research assistant, using LangGraph — helping users cut through massive datasets to surface precise, verifiable investment insights.
— LangChain (@LangChain) 17 juin 2025
Aisis Julian (Senior Software Engineer) and her team at Morningstar have… pic.twitter.com/jY661Ze4vcInvestment research, accelerated. @MorningstarInc built Mo, their AI research assistant, using LangGraph — helping users cut through massive datasets to surface precise, verifiable investment insights. Aisis Julian (Senior Software Engineer) and her team at Morningstar have
-
Testing AI prompts is as essential as testing code
By
–
Testing isn’t optional anymore. It’s table stakes for any serious AI product. Your prompts are the foundation of your product. Test them like you test your code.
-
Why prompts need testing across hundreds of inputs
By
–
Your prompts need to work across hundreds of different user inputs. One good example doesn't guarantee that your prompt will handle edge cases effectively. Without proper testing, you’re shipping prompts that work sometimes. That’s not good enough for real products.
-
Prompt Quality is Key to Reliable AI Products and User Satisfaction
By
–
As AI agents become more complex, the quality of prompts directly impacts the reliability of your AI product. Bad prompts = unreliable AI = frustrated users.
-
LLM Weaknesses: Making Large Language Models Practical
By
–
LLMs are not plug-and-play geniuses; they often need extra work to be practical in real-world applications. That's what this week's iteration is about: https://
open.substack.com/pub/louisbouch
ard/p/llm-weaknesses-101-what-they-really?r=25qlky&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
…