lmeyerov

19 ms·

271.
by lmeyerov 2y ago
Half our work is on bringing GPUs/graphs/AI/etc to Serious Problems like cyber, misinfo, fraud, crime, etc, and a few things I've come to: * If you're unsure, quit * It's fine to distrust the system, and as wit
272.
by lmeyerov 2y ago
Super cool. Any way to know if the endpoints are 'ethically sourced' and how they are secured wrt confidentiality & integrity? (We have a 100K+/day scraping workload, and TBD full interactive automation)
273.
by lmeyerov 2y ago
For security & fraud teams who want to 'own their AI' vs trust with Sam Altman, we are doing some fun things here as part of louie.ai, and looking for our next cohort of Splunk/databricks/elastic/neo4j/etc
274.
by lmeyerov 2y ago
Moving from langchain interop to protocol interop for tools is great Curious: 1. Authentication and authorization is left as a TODO: what is the thinking, as that is necessary for most use? 2. Ultimately, what does MCP already add or will a
275.
by lmeyerov 2y ago
'Irregular' computations are historically a challenge for GPU acceleration, so seeing this high initial debut of the open source project is interesting We are long-term cugraph users for pygraphistry and the newer GFQL graph query
276.
by lmeyerov 2y ago
Hah indeed, we are doing billion-scale real-time graph rag in louie.ai for fairly regular tasks, so your sentiment resonates ;-) For something like uploading a big folder of documents, agree with the OP, pretty straightforward, naive in-mem
277.
by lmeyerov 2y ago
Always a fan of query plan articles! Note: the dig at dataframe libs is worth some care in case you think that means duckdb can optimize and they cannot Dask, Polars, and others pick a lazy default in order to make distribution and other op
278.
by lmeyerov 2y ago
We are looking for something to run as part of our ami/docker testing and as you say, stays fresh on standards (whatever soc2/iso, but ideally also FIPS) , any prefs?
279.
by lmeyerov 2y ago
Thanks! AFAICT, a lot of that comes down to storage abstractions, which I'll be curious to see the answer on! Pinned localstorage <> cloud native is frustrating. I sense another big chunk is the fast secure start problems that fi
280.
by lmeyerov 2y ago
I was intrigued because the development environment problem is similar to the data scientist one - data gravity, GPU sharing, etc - but I'm confused on the solution? Oddly, I left with a funny alternate takeaway: One by one, their clev
281.
by lmeyerov 2y ago
I agree with the manager-knows mentality for people management evaluations of productivity focused on bonuses & firing... they should know! ... ... but disagree from the high-performing engineering org notion of individual & team pr
282.
by lmeyerov 2y ago
Graphistry, Louie.AI | Full stack, security AI evangelist, devrel, growth marketing | REMOTE GLOBAL | https://www.graphistry.com/careers The Graphistry GPU-accelerated visual graph AI platform is used by data-intensive inve
283.
by lmeyerov 2y ago
It's hard to square seeing the magnitudes of capability coming from genAI & AR and thinking the web UX is done because someone's idea of browser tab layout is boring How we use the web today, with the Chrome engine and skins a
284.
by lmeyerov 2y ago
Yep. 1-2 years is a judge. The official median is 7 years, which afaict is being quite generous. And agreed, it's gotten a lot harder even for folks most would say we should be fighting to attract. One of the best things the US can do
285.
by lmeyerov 2y ago
I was excited for that headline, but I didn't get a clear & fair sense of comparison, like how prompt engineered etc was the comparison, or a false comparison? There was not enough detail to determine what was ultimately useful &am
286.
by lmeyerov 2y ago
We see them for smaller transaction extracts, but most vendors here are too expensive for most teams at the scale you are likely thinking. Instead, we see log, kv, dwh, etc systems -- think elastic, Clickhouse, databricks, and the like. Som
287.
by lmeyerov 2y ago
I encourage you to run evals on result quality for real b2b tasks before making these claims. Almost all of your post is measurably wrong in ways that cause customers to churn an AI product same-day.
288.
by lmeyerov 2y ago
Pretty good - r=2 is a regular graph afaict, and basically anything that maps to a frontier-based pattern works well. Ex: level synchronous bfs during topological sort. For the 'easy' way we do in gfql, which is basically vector o
289.
by lmeyerov 2y ago
Nice! It's interesting from the perspective of maintenance too. You can bet most constants like warp sizes will change, so you get into things like having profiles, autotuners, or not sweating the small stuff. We went more extreme, and
290.
by lmeyerov 2y ago
You use their APIs in a way that commoditizes them. Ideally your customers don't care if you switch to Anthropic, because the LLM provider is not the reason customers are picking you. Likewise, there is some structural reason that Open
291.
by lmeyerov 2y ago
At least for Louie.ai, basically genAI-native computational notebooks, where operational analysts ask for intensive analytics tasks for like pulling Splunk/Databricks/neo4j data, getting it wrangled in some runtime, cluster/
292.
by lmeyerov 2y ago
IMO that kind of thinking flips correlation/causation, and feels like some sort salience bias towards voluminous but crappy work. Yes, you can do it, just good bye impact. Worse, it's a self-fulfilling prophecy: by moving your w
293.
by lmeyerov 2y ago
As I wrote, different folks will already be chasing more immediate projects as they're too attractive: Think masters students, industry labs, startups, etc. Rules are meant to be broken. For example, a professor will often put a mas
294.
by lmeyerov 2y ago
This is fun. At a fractal level, it's what's going on in graph RAG, where you're doing hierarchical similarity modeling and using that structure in different ways for smarter summarization, indexing, & retrieval.
295.
by lmeyerov 2y ago
I like the article directionally but fear the examples are too short-sighted for most AI researchers. Picking something useful in 1-2 years is a reason to go to industry, not research, and leads to mostly incremental units that if you don&#
296.
by lmeyerov 2y ago
Ah yeah unfortunate, afaict it does not (yet?) support lambdas, which is where imo this starts supporting flows that are hard via manual asserts I was hoping something like a more modern https://dl.acm.org/doi/10.1145&#
297.
by lmeyerov 2y ago
One of the most 'nice' thing about contract systems is automating blame checking in higher-order code. They handle the lambda/method param wrapping and blame direction assignment for you, which is annoying to do via manual as
298.
by lmeyerov 2y ago
Congrats, definitely an area we're always examining tools Afaict, this seems like a more restrictively licensed & less capable alternative to Dagster and Prefect. However, there may be some specific areas it is ahead -- the multite
299.
by lmeyerov 2y ago
This was my exact question. Why do an LLM rewrite, when you can add a context vector to a chunk vector, and for plaintext indexing, add a context string (eg, tfidf)? The article claimed other context augmentation fails, and that you are bet
300.
by lmeyerov 2y ago
We are fans! We mostly use cudf/cuml/cugraph (GPU dataframes etc) in the pygraphistry ecosystem, and when things get a bit tricky, cupy is one of the main escape hatches

More ›