Currently thinking about evals and benchmarks... distributedthoughts.org/the-benchmar...
distributedthoughts.org
Abe Gong
@abegong.com
geek dad with a clipboard founder @ Katabase and Great Expectations operating advisor @ Bessemer VP data, ai, startups, dev tools, systems thinking storytelling, community, creativity blog.abegong.com
Currently thinking about evals and benchmarks... distributedthoughts.org/the-benchmar...
distributedthoughts.org
So.... it seems like now we can post about AI on bluesky without getting violently mobbed. ... yes?
Before my partner goes camping, she counts the exact number of marshmallows and graham crackers she plans to use for s'mores. I don't hink I could live that way
Claude Code, trying to get that one last unit test to pass
incredible things are happening in China
Inside you there are two wolves. One is active, the other one is on hot standby and becomes active if the first one fails or is taken down for maintenance. Add more wolves as necessary for increased redundancy load balancing. A quorum badger can be added for environments with multiple active wolves.
Looking forward to Data Council this week! HMU if you're around and want to catch up.
FastMCP 2.0 is out! New features include: 🔄 Proxying any existing MCP server as a FastMCP server 🧩 Composing multiple servers into one 🪄 Autogenerating MCP servers from OpenAPI or FastAPI 🧠 Client LLM sampling Let's keep growing the MCP ecosystem! 🚀 github.com/jlowin/fastmcp
GitHub - jlowin/fastmcp: The fast, Pythonic way to build Model Context Protocol servers 🚀
The fast, Pythonic way to build Model Context Protocol servers 🚀 - GitHub - jlowin/fastmcp: The fast, Pythonic way to build Model Context Protocol servers 🚀
github.com
Last year we encountered a strange glitch in one of our surveys…the word “yes” was being replaced with the word “forks” in response options. Read on to learn how we unraveled what was causing it, how we ensured it didn’t impact the data quality, and what to do if it happens to you!
How a glitch in an online survey replaced the word ‘yes’ with ‘forks’
Dating back to at least early 2023, a bizarre and alarming technical glitch started popping up in some organizations’ online surveys and forms.
pewresearch.org
I'm looking forward to digging into MongoDB's new AI showcase repo: github.com/mongodb-deve...
GitHub - mongodb-developer/GenAI-Showcase: GenAI Cookbook
GenAI Cookbook. Contribute to mongodb-developer/GenAI-Showcase development by creating an account on GitHub.
github.com
Can't stop giggling at this sample question used to test an AI reasoning model in a paper I'm reading
The pandemic started five years ago, which means it's just about time for Cisco to announce big product enhancements to WebEx to help cope with the social distancing.
In the last few days, I'm finding myself fighting memories in Windsurf a lot more often. Cascade will often get hung up on irrelevant/outdated info in memories and ignore/garble specific instructions. It started happening right after I installed the Wave 4 updates. Anybody else seeing this?
personally would not describe the myth of a man cursed to roll a rock uphill FRUITLESSLY FOR ALL ETERNITY as “a reminder that you can overcome anything” but you do you, meta ad
OMG, Windsurf/Claude is amazing at taking the pain out of configuring Github actions.
Ugh. For reasons I cannot fathom, I woke up with Jack Black/Bowser's "Peaches" song stuck in my head. Thanks, Obama.
This datavis and the layout algorithms behind it is a tidy little thing of beauty
For the German election we wanted to use a cartogram of German election districts but didn't find anything good, so we created our own. Here's a little nerd thread how the final algorithm worked 🤓 🧵 #gis #cartogram #election #datavis #cartography
Are there any @pydantic.dev power users on bsky? I have a Question: github.com/pydantic/pyd...
What's the recommended way to implement managed attributes in Pydantic v2? · pydantic pydantic · Discussion #11493
I have a use case where I need to implement managed attributes in my Pydantic models, with the following requirements: The attribute should be settable through the constructor After construction, t...
github.com
Is there a name for pre-refactor "I'm not sure yet how bad it's going to be" anxiety?
Nothing says "this software isn't for me" quite like having to give anything a "campaign name." If I'm forced to give campaign names in a tool I always call it "winter land war in Asia"