lmeyerov.bsky.social - Profile | ThreadSky | a Reddit-style client for Bluesky

Curious how startup folks approach release velocity nowadays. We're revisiting gitflow assumptions: - localdev - dev branch <> server - big changes on long branches - SaaS: most devs deploy 1-3x/week - Enterprise: 2w-4w cycle, with customers updating quarterly keep in mind we're < 20 ~senior devs

submitted 7 days ago • 0 comments

Sunday funday: Make git repos friendly for LLM use with a tiny bash script - (sharing from some internal louie_ai tooling) - filters & excludes for desired text files - convert to LLM-friendly markdown that won't be confused with instruction prompts - special casing of ipynb - multicore gist =>

submitted 12 days ago • 1 comment

Graphistry won the US Cyber Command AI competition by auto-correlating alerts into incidents and kill chains. Our graph ML clusters alerts fast and links them into a timeline. CEO Leo Meyerovich demos our one-node GPU solution processing over a billion alerts. #cybersecurity #hunt #SOC

submitted 21 days ago • 0 comments

Something the louie team is seeing that I haven't seen discussed: Reasoning LLM fine-tuning for agents & reasoners * Ex: Multiturn quality drops when reasoning LLM's effective CoT nests with CoT agents and get confused * Ex: Domain specific scenarios like legal & medical needing specific guidance

submitted 24 days ago • 1 comment

How AI is transforming in general seems better understood by zooming out to see AI as fuel on a much bigger fire that has been growing the last ~30 years: - The internet won: 5.5B users - Software ate the world: DBs+APIs for everything - AI software 2.0 is here: Small teams moving faster

submitted 28 days ago • 1 comment

farewell WYSIWYG, hello WYPIWYG: What You Prompt Is What You Get

submitted 38 days ago • 0 comments

Here we go: @graphistry.bsky.social / louie.ai + Google Spanner Graph team phase 1 Best visual graph intelligence experience + one of the best databases ever made Congrats team, and excited for the next phase Happy graphing :)

submitted 41 days ago • 0 comments

LLM redditor replacement: If 'expert agreement' measures whether experts like an LLM answer.. Then 'redditor replacement' is whether experts prefer the LLM over a random redditor Context: Our users eval us expert-level quality, but benefit as soon as they can trust the LLM more than a redditor

submitted 47 days ago • 0 comments

We are working on more how-to's split for new folks vs advanced, and the first is fun as it describes the 80% case for starting graph analytics & viz projects: turn any table into a graph and go If any more that would be useful, lmk!

submitted 48 days ago • 0 comments

And @graphistry.bsky.social is up! Starting with a bang: Video+slides of our Suricata #suricon24 collab w/ Anthony Tellez @ BNY This gets into meat-and-potatoes of genAI for rules & IDS pipelines, and what to do when RAG workloads get big .. Also always fun live demoing louie.ai & graphistry 🤠

submitted 49 days ago • 0 comments

Finally looked it up (Or: Why I'm messing with dataframe column order in our hands-free feature engineering library on a Saturday morning)

submitted 55 days ago • 0 comments

Been fascinating whitebox AI red teaming #loiue_ai (graphistry 2.0). Four 🤯 areas: - Agent attack painful for businesses - New prompt injection nasty to data teams - Embeddings attack - China Month+ for us to digest: I want to write up. Recommend for all real AI product teams!

submitted 55 days ago • 0 comments

Very cool to be saying... Graphistry / #LOUIE_AI is looking for ~3 R&D folks at Director + Fellow levels. Join as we build out our OSS GPU graph query language GFQL, our autonomous investigator, and formalize our R&D subdivision to run special projects in general!

submitted 56 days ago • 0 comments

1/ $B AI coder fork-in-the-road: - Wired: LLM generates & passes test cases - Inspired: LLM generates & passes proofs Ex: * python => typed python => rust => z3/coq/f#? * real UX to guide proof conditions & solvers?

submitted 64 days ago • 1 comment

Say hello to the Graphistry 2025 GPU AI Remote Fellowship program! 1/ We're looking for 2 student researchers (PhD/MS/BS) to hack on: 1. Autonomous AI analysts 2. Multi-GPU OSS graph query languages -->

submitted 72 days ago • 1 comment

Was chatting on indirect prompt injections (second-order + poisoning), and my shifted stance: 1. Most discussions miss the boat: Sabotaging your own session or stealing a non-secret system prompt is typically a low risk. =>

submitted 73 days ago • 1 comment

The louie.aim team is working through our first AI red teaming, and curious if anything folks found useful? For round 1, we're targeting agentic systems & genAI data pipelines, both manual+auto attacks, vs the model & supply chain

submitted 74 days ago • 0 comments

While some AI researchers complain about being GPU poor, I think it's an incredible time to be doing research, so many methodological doors have opened for most fields! Good thread!

submitted 76 days ago • 0 comments

peak california

submitted 76 days ago • 0 comments

Neat experiment: We did a paid code challenge on a recent Graphistry / louie.ai role $100 or a donation for first 10 passing submissions on a 1 hour task - 100+ CVs, not all submitted - 6 winners -> donors - 3 winners -> $ - 1 winner -> 50/50 Result: All short-listed CVs were in the 10.. & donors

submitted 76 days ago • 0 comments

US vs Asia mask culture post-COVID converging? I got a cold on a recent work trip, so put on a mask on flight back out of courtesy. While normal elsewhere, rare in US pre-COVID. 10+ on the plane did the same, so the COVID norm continues! However: Unsure of % as courtesy (Asia) vs for self

submitted 78 days ago • 0 comments

Curious: How are folks finding llama 3.3 70b in practice, e.g., are they finding it to be between 4 and 4o, or where would you peg it? And especially for summarization/extraction, reasoning (CoT), and coding/querying?

submitted 80 days ago • 0 comments

This is a wonderful look at just some of the work that goes into a good text2query system, or neurosymbolic genAI agents in operational systems in general. Likewise, it also hints at why simply training an LLM won't solve all your quality problems: www.linkedin.com/blog/enginee...

submitted 80 days ago • 0 comments

Thought this visceral reaction was worth sharing, definitely a thing for those building LLM-powered tools that I talk with and our own team www.linkedin.com/posts/leo-me...

submitted 83 days ago • 0 comments

ODNI's new standard aiming for lightweight citing genAI intel in reproducible ways is interesting directionally. Streamlining and scaling flows ia a tenet of louie.ai and Graphistry design - court admissibility - so seeing where ODNI goes pragramatic is interesting www.linkedin.com/posts/odni_i...

submitted 86 days ago • 0 comments

Any others like this for security, graph, misinfo, or other areas folks would recommend?

submitted 91 days ago • 0 comments