I found this page on OpenAI's help documentation very timely and useful:
help.openai.com/en/articles/...

A year of juggling search and AI I've now spent just over a year working on my own since leaving OpenSource Connections - it's been a time of constant challenges and opportunities both work-related and personal. I started the year by...

A year of juggling, search & AI www.linkedin.com/pulse/year-j...

Just had to provide a photo ID to a third party verification site to use OpenAI. I’m not looking forward to this trend.

The CFP for Haystack EU is now open!

opensourceconnections.com/haystack-eu-...

If you're using a coding assistant like Cline, it's incredibly useful to have it keep its own notes about the context of the task and any important things it's learned along the way. I've been keeping cline_todo.md and a cline_notes.md files in my projects and it saves a lot of time (and tokens).

OpenRouter The unified interface for LLMs. Find the best models & prices for your prompts

Go ahead and give openrouter.ai a try. So far my experience has been nothing but positive and it's hella cheaper than what I was using before.

I've never hated documentation as much as I hate GCloud's documentation. I DARE you to implement a solution just based off of their documentation.

Ok, Amazon Q was moderately helpful to me today, but I'm scared to death I'm going to be in the middle of a project and run out of free tokens. It was a pleasant surprise that it had agentic access to my resources and could inspect them.

I’m still grinding at the satellite tournaments hoping to win a seat

Beyond LLM-as-judge towards LLM relevance engineering Most now agree that LLMs can provide useful search judgments. In this talk, we’ll discuss the next level we’ve achieved at fashion search startup Daydream: a...

In the vibe of cheating at search with LLMs...

I'll be speaking at Haystack next week on LLM-driven relevance engineering. Join me!

haystackconf.com/us2025/talk-8/

I've been really impressed with the way Cline works with vs code's terminal, but there are definitely times when I need to coach it on ways to get to what it wants without generating a bunch of tokens to read.

Haystack - The Search Relevance Conference Haystack is the no-holds-barred conference for practitioners working on search, matching, and relevance.

Come join 150 of the most awesome folks building amazing search to figure out this changing landscape at Haystack! haystackconf.com. Early bird tickets through Sunday. Come early for Haystack and attend PyData Virginia! 2 conferences in one week! ti.to/pydata/pydat.... BBQ at my house too!

Okay, smirk all you want at "vibe" coding, but it's more productive than duck debugging and much more available than live pairing.

Nice! Apparently there's an expert technique of asking indirect questions when you can't or don't want to ask the direct question. I wonder how well ChatGPT (et al) can do that.

This is Marco Rubio explaining how the USA promised to defend Ukraine forever if they got rid of their nuclear arsenal left after the Soviet Union fell.

This is why lil marco was sinking into the couch. He was hoping we wouldn’t find it…so don’t RT right now this very second.

One of my German colleagues suggests this instead: Vergangenheitssehnsuchtssong - yearning-for-the-past song

Vergangenheit ("past")
Klang ("sound" or "tone")
Sehnsucht ("longing" or "yearning")

I'm inventing a new word in the German fashion:

Vergangenheitsklangsehnsucht - Bittersweet yearning or nostalgia triggered specifically by hearing music from one's past.

I really do hope that our quest to root out hallucinations in LLMs helps us fight _human_ generated BS

Today I was also reminded that there’s no “easy” substitute for relevance.

We did some math to work out what the cost to roll out vector search would be… we’d need to increase our compute budget nearly 50%…. Just to roll out semantic search for issues.

Just discovered something more daredevil than live coding during a presentation: Letting your coding agent live code.

Wow, this is a first for me and I can't even fathom how hard this was to implement. If you use OpenAI's v1 API but you have v2 installed it will give you an error saying you can do "openai migrate" to automatically migrate your code. And it works!

Someone called it this generation's Sputnik moment. Maybe a bit of hyperbole, but same kind of feel

Hi Max!

eh? I got the notification

Anyone read that article about "the third place"? I kinda feel like that's bsky. It's not work (LinkedIn), it's not home (Fbook), it's just... 🤷

Excellent point.

Highlighting. You _can_ do this with an external service but it's non-trivial. Most folks would rather just pay for the extra storage.