nwatson

Sounds interesting. Now go make me a cappuccino before my next ride.

I agree. If grandma grows her own tomatoes she can only try four tomato-related recipes. If she goes to the grocery store she can try thirty.

The AI is enabling a lot more experimentation with software. The AI supplies the tomatoes (infrastructure / grunt work), grandma tries her recipes (the software / business idea).

The choir director at my church, Michael Dodds (UNCSA), covers the early development of musical notation, "From Modes to Keys in Early Modern Music Theory" [0]. Currently on sabbatical at Cambridge. Once in a while during rehearsal he will give an impromptu lecture on music history, always interesting.

[0] https://a.co/d/0bLcrlyE

In Brazil in the 1980s the VASP (Viação Aérea São Paulo) airline put their smoking section on the right side of the single aisle, and the non-smoking on the left side of the single-aisle. The smoke stayed on the right side just fine. (jk)

São Paulo is in the "green" but they were supposedly around two weeks from running out of water for municipal distribution in 2014/2015. A city of that size without water would not do well.

https://en.wikipedia.org/wiki/2014%E2%80%932017_Brazilian_dr...

https://www.smartcitiesdive.com/ex/sustainablecitiescollecti...

EDIT: added timeframe

Did they airlift better supplies to a floating aircraft carrier somewhere in the northern Indian Ocean recently? No.

One of my jobs at work is to translate data-science-provided AI models, convert them to optimal GPU forms, and send them downstream to DevOps for deployment in a variety of contexts, all contexts adaptable from basic Docker instructions.

I had a couple of very specialized adapters for the first couple of environments that were very aware of the whole GPU conversion and deployment stack (ONNX, TensorRT, Triton, etc.), but even these were very fragile to version upgrades, etc., and required a lot of adaptation from one "stack" to the other. As soon as the frontier models got to be real good I took a more "agnostic" approach. My new framework's goal was to be "mealleable" and very hands-off w.r.t. the details of the conversion / deployment ... after all, these models are very well trained on the whole AI pipeline, including deployment. So now the basis is more or less (a) where are the model files, which Docker image do we want to use, which "conversion / deployment" method(s) do we want to try, what are the inference use cases addressed; (b) what is the conversion / deployment method used? (c) how do we assemble the smoke-test and basic performance test cases for multi-client, single- and batch-processing modes for all use cases? (d) how do we package the end results so that the DevOps person has all they need to make sure they have the proper files and that they can run Docker and check that the inference works? These questions and answers are all wrapped in some very flexible base classes. Every conversion is somewhat different, so each one involves extensive discussions with Claude Code (e.g., "hey, look at this other prior conversion, its raw files, compare with these raw files, develop a conversion / use-case-smoke-and-performance-test, and packaging). Rather than trying to be very hands-on at these lower level, Claude has almost free reign at the "low level" to suggest the best approach, and I so I "talk to an engineer with vast knowledge but (for now) a bit less judgment". With this method though I probably cut the total time to conversion-for-deployment by 80% to 90% now ... the choices and options are vast, and Claude knows a lot more than I do.

At times there is not a ready-made solution for the particular problem at hand, and so then it gets more interesting with how to shoehorn a solution into one of the available technologies. We prefer one of the various builds of the Triton Inference Server (or one of its hardened versions maintained by others).

As incomplete as the sentences from one of these may be, their "weave" style tickles the attention mechanisms for their most ardent supporters, and comes across as heartfelt, relatable, and ultimately a coherent stream of consciousness dump.

Some companies hire junior devs to adopt AI-first practices to then drag their seniors in the same direction.

I'm hopeful that software will be applied in new areas where it was too costly before, with AI help it will be possible. Software will be cheaper, even more prevalent, lower quality in since respects, but SW engineering will still be a viable career.

We better hope there aren't cuts to -rabies wild animal vaccines. These have cut down incidents like these in large degree, https://www.aphis.usda.gov/national-wildlife-programs/rabies... .

QUOTE:

The National Rabies Management Program was established in recognition of the changing scope of rabies. The goal of the program is to prevent the further spread of wildlife rabies and eventually eliminate terrestrial rabies in the United States through an integrated program that involves the use of oral rabies vaccination targeting wild animals.

Since 1995, Wildlife Services (WS) has been working cooperatively with local, State, and Federal governments, universities and other partners to address this public health problem by distributing oral rabies vaccination (ORV) baits in targeted areas. This cooperative program targets the raccoon variant, canine variant in coyotes and a unique variant of gray fox rabies.

I've searched out this clip various times, gonna watch it again. Easily one of the best movie / TV moments since the advent of film.

Health insurance company margins are 3% to 5%. The book America’s Bitter Pill: Money, Politics, Back-Room Deals, and the Fight to Fix Our Broken Healthcare System (2015) by journalist Steven Brill blames mostly wasteful practices in hospitals, pharma and device manufacturers, government and corporate health administration costs.

One of my tasks is to take various AI models the data science team has produced/fine-tuned and make them runnable with optimizations on GPUs (TensorRT, vLLM, Triton Inference Server, etc.), involving conversion, deploy, smoke-test, packaging, documentation, all within a uniform framework. The documentation includes model-provenance, simple deployment instructions (usually a Docker run), simple deployment test. The DevOps team takes this and tailors it to Kubernetes / helm-charts or whatever the target environment needs.

I had worked on some earlier deployment environments for a few models where we focused on one version of Triton and one runtime technology. Even upgrading to a newer Triton version was brittle, involving a lot of command-line changes at various phases. This was written mostly pre-Claude-getting-real-good. I decided that probably Claude had matured and was way better at understanding the particulars of AI-model-GPU-deployment-and-technologies. I worked with Claude to make the framework a much more lightweight wrapper, ignorant for the most part of a lot of the deployment internals.

After this refactoring and doing the first couple of models, it's quite amazing at how well Claude can figure things out. For any new model we now set up its "specialization" directory and its documentation subdirectory, point Claude at the proper AI-model files, point Claude to the sample non-optimized inference code and test data, and point Claude to a similar conversion we've done in the past. There a multi-layer class hierarchy dealing with various tiers. I ask Claude to explore the existing conversion/packaging, the model, any documentation that comes with it (a lot of times there are unexpected twists), the sample code, and the desired multiple use cases the model is meant to address, and the test data. Claude has been trained, I'm sure, on a lot of AI model conversion, so it's able to synthesize the full multi-stage conversion/deployment pipeline, come up with appropriate test cases for all the use cases involved. There are usually between 3 and 10 refinements after initial synthesis, fixing outright errors, refinemnts that the data-science team requests after playing with test deployments, etc. The options and pitfalls are vast, and without Claude each preparation likely would 10x or more longer. I just put most details in Claude's hands, and make sure the general framework is good enough to provide external uniformity. When all is working, it takes a couple of hours to make sure the documentation is good.

All this to say that at least for this domain, Claude / AI has been a game-changer and has sped up the process amazingly.

Not quite one note, but "Open Eye Signal" (2023 remaster) by Jon Hopkins has wide spectrum background and heavy rhythms, but once the song gets going the main instrument moves between fifth -> seventh-flat -> sixth -> fifth on repeat, and that's the melody for the song. And yet, to me, it's captivating.

The boiled water (assuming even heating throughout) is what reaches boiling temperature, 100°C, which is the entire pot. Additional energy applied to the pot/water after that point turns at some rate the already-at-100°C-boiled-water into water vapor/gas/steam. If you stop heating before all the already-at-100°C-boiled-water disappear, you have left what is or will have been "boiled water".

I opine that over-hyphenation adds clarity.

Joe Frank ... some of the best audio out there. I wonder whether one had to grow up at a certain time for his shows to make sense, and whether they'd be lost on newer generations.

Cost goes down significantly for subsequent years personal. Currently at less than $200 yearly.

I'm curious, did you choose "blind pilot" as your username before or after your adoption of LLMs for projects?

When I bought one they had "reader" lenses that were not prescription, they did well enough. I use 2.0 diopter reading glasses otherwise from Costco.

I was in the Winston-Salem Flywheel coworking for eight years, good place. I hope the Greenvile site is as good.

I have arthritis in my right ankle now and walking more than a mile leads to swelling, pain, and random sharp pinching during some strides. Running sets it off sooner than that. The ankle always is visibly swollen, it's obvious there's something wrong, and after more acute symptoms developed about eight months ago after my first soccer action in about a year I finally got it checked out and diagnosed two weeks ago. I'm lucky in that I can still cycle for an hour or two on fairly hilly roads in my North Carolina town, and I've been enjoying riding in the Phoenix desert a few times while visiting family this past week. I can still do some weight-bearing exercises too.

My biggest regret though is that I may never manage to play more than a few minutes of soccer at a time again. I got back to Latin America in early adolescence having missed some crucial soccer years. I was soon a couple of years younger than everyone else in my grade, and P.E. classes were not very fun, it was hard to compete and I rarely got to participate in real action on the soccer or rugby field. In my late teens I started to actually develop some soccer sense and got a bit better. But student/teacher political strikes during the dying years of a dictatorship and upcoming return of my family to the USA brought me to the USA for studies, and I didn't play much in college.

After a few years in SF Bay Area I started playing pickup soccer and eventually got to play quite well , especially during a particular two year stretch. Then marriage, busy jobs, having a kid meant I laid off the regular soccer for a while.

And now, with a bit more extra time I could maybe spend playing I no longer can. I've never been on a team, never been a specialist at a position, never trained regularly. The doctor said maybe with physical therapy and pain killers I could do it. I'll work toward that.

Medium / large companies won't take the risk on smaller operations selling a new focused tool unless it's a major pain point. They'll pay more for less risk, assuming the PE-managed company will go out of their way with account management to address all their concerns.

AVGO/Broadcom in some way acts like a big PE firm, rolling up other software companies, integrating them into their huge suite of offerings, ousting the new integrated offering's competing tools from the customers environments and selling the increment, and cutting off smaller customers not willing to subscribe to the huge suite.

Equivalent PIX is very popular in Brazil for instant payments.

In letter and spirit, yes. If you're in the club, no.

It's probably harder to model and the results "aren't quite there yet".

Nate B Jones videos ... YouTube channel "AI News and Strategy Daily" channel uses all of these. Every video.

Admissible evidence probably requires parallel construction and then a warrant. The purchased data is the catalyst but not legally actionable.

If the Claude (or similar) can't open a browser on a headless server, they typically print a URL you can copy to your browser on your local system-with-GUI. From there you authenticate and get back some kind of token, which you copy and paste back into your remote SSH session.