lmeyerov

40 ms·

481.
by lmeyerov 3y ago
This appears to be the official account, not a personal one. If he wants me to take htmx seriously as a prod dependency, I care about how he treats it, all the way from code to governance to community interactions. He has no such obligation
482.
by lmeyerov 3y ago
may you and the people relying on you never have to rely on the maintainers taking things more seriously than they've shown themselves to in their public positioning
483.
by lmeyerov 3y ago
may you never have to maintain production code...
484.
by lmeyerov 3y ago
I think 'environments' was meant, where diff GHA environments get diff secrets, and policies dictate who gets to run what actions with what envs. But that is real work to setup, audit, and maintain. It'd be better if, like ph
485.
by lmeyerov 3y ago
Agreed. Agent reasoning systems should learn based on past and future use, and both end users and maintainers should have power in how they work. So projects naturally progress on adding guard rails, heuristics, policies, customization, etc
486.
by lmeyerov 3y ago
Agreed, maybe I wasn't clear enough. I don't view it as BI team vs platform team vs whoever. Maybe a decrease in the need for PhD AI consultants for small projects, or to wait for some privileged IT team for basic tasks, so they
487.
by lmeyerov 3y ago
Those are pretty normal needs for us and our users. A big reason louie.ai exists is to make easier all the years of Graphistry helping enterprise & gov teams use python notebooks, streamlit/databricks/plotly python dashboards,
488.
by lmeyerov 3y ago
Yep. A lot more on our roadmap, but a lot already in place! It's been cool seeing how different pieces add up together and how gov/enterprise teams push us. While there are some surprising implementation details, a lot has been fo
489.
by lmeyerov 3y ago
Our theory is we are having simultaneously a bit of a Google moment and a Tableau moment. There is à lot more discovery & work to pull it off, but the dam has been broken. It's been am exciting time to work through with our custom
490.
by lmeyerov 3y ago
We have been piloting louie.ai with some fairly heavy orgs that may be relevant: Cybersecurity incident responders, natural disaster management, insurance fraud, and starting more regular commercial analytics (click streams, ...) A bit unus
491.
by lmeyerov 3y ago
It's great in prod. Doesn't make $ for companies marketing k8s iaac devops, so no one to advertise.
492.
by lmeyerov 3y ago
Oof. For visualization engineers impacted, we have a global remote position for our GPU server <> GPU client engine, working on gen2 scaling for the next 100X of our award-winning infoviz platform: Python/RAPIDS <= Apache Arro
493.
by lmeyerov 3y ago
Important point. Nowadays, we write a pure pandas version, and when the data needs to be 100X bigger and faster, change almost nothing and have it run on the GPU via cudf, a GPU runtime that fully follows the pandas API. Most recently, we p
494.
by lmeyerov 3y ago
model reverse engineering is a pretty cool research area, and one big area of it is figuring out the training sets :) this has been useful for detecting when modelers include benchmark eval sets in their training data (!), but can also be
495.
by lmeyerov 3y ago
AI leadership seems existential to being a top data management company and providing top data management capabilities: * Databricks data management innovations, now that basics are in, are half on the AI side, like adding vector & LLM
496.
by lmeyerov 3y ago
That's a fair surface-level view, but worth thinking through a bit. Palantir is multiple main things, and a whole ton of custom software projects on top, and a good chunk of them rely on the quality of their NLP & vision systems fo
497.
by lmeyerov 3y ago
Interesting, thank you for sharing! That experience speaks more to the perils of in-housing, not to why Palantir is the best COTS for specific needs here. Are there specific leading COTS here you view it so far ahead of for such a contract?
498.
by lmeyerov 3y ago
The acceptance of search engines set a lot of precedence that make any blanket reasoning harder to justify - the underlying technologies and use cases are often similar or even the same. Google has been using both extractive and generative
499.
by lmeyerov 3y ago
Graphistry | Viz engineer | Security engineer | Contract | REMOTE - GLOBAL | https://www.graphistry.com + https://www.louie.ai Graphistry is a Gartner-awarded startup used by fraud/cyber/BI/supplychain
500.
by lmeyerov 3y ago
This is a great article and fun to see fundamental concepts get (re)discovered here! A perspective that we can generalize from the hierarchy discussion is to think about tool-for-the-job: what is the 'content' job, and what '
501.
by lmeyerov 3y ago
Not running eval on third party model weights when encouraging consumers to download them seems like the low bar that comes after have any non-executable policy at all, especially for something Mozilla supported. Edit: I mean as the default
502.
by lmeyerov 3y ago
We are trying to saturate storage->pcie->gpu cards for tasks like gpu-accelerated log analytics, and this is increasingly the bottleneck
503.
by lmeyerov 3y ago
We have been excited to dig into the Iceberg era of more managed parquet storage... But they are still years behind on supporting fast GPU IO (GPUDirect/cuFile). So every time we look at bringing them to a customer for powering AI work
504.
by lmeyerov 3y ago
KG in the intuitive 90's sense is fundamentally worse, but with llm-era rethinking, potentially useful - kg DBs have the same retrieval scale problem, and often built on top of the same DBs - traditional kg mining would still use LLMs,
505.
by lmeyerov 3y ago
I like this article A tougher question for me here is the why, vs how: 'Why research LLMs in academia if you can work in industry on the same problems with more resources?'. Systems researchers face something similar, where big sy
506.
by lmeyerov 3y ago
Specifying networks is one of the main docker compose commands and seems pretty normal: services: your_service: image: your_image networks: isolated_network: another_service: image: another_image
507.
by lmeyerov 3y ago
The FirefoxOS scenario would be Gecko embedding in FirefoxOS, so Mozilla is their own customer, and makes $ by being the controller of FirefoxOS - OS search, and configuration of browser search It's less likely the handset manufacturer
508.
by lmeyerov 3y ago
A couple agencies having certainly been flexing like that The chilling effect may be more around the design tools market specifically The case of Adobe is a bit different than anti-monopoly movements against other $T because the lack of &#
509.
by lmeyerov 3y ago
Interesting to work through that On its face, it's a tough sustainability position for an org of Mozilla's size b/c browser revenue - today - is from controlling the search bar (and maybe now payments APIs?). By making Firefo
510.
by lmeyerov 3y ago
Awesome to see. We are productionizing a variant of this in louie.ai and it's the biggest step function in system quality we've seen in ~months. This kind of thing enables continuous learning & user-directed learning without h

More ›