This works: https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fgithub.com%2FSciSharp%2FLL... Using Ollama to dump clean models I built a c# app using Llamasharp in a couple of days that cycles history (.. learns?), saves sessions, runs offline, etc. albeit slowly on my 2080. I'm WAY too cheap to pay for ChatGPT, and I wanted to better understand what Deepseek was so I've been using the 14b model as a baseline, and it works great. It can refactor itself (feed forms/documents into it), and writes some pretty mean Godzilla poetry in XML format when prompted. So to recap, Ollama is fine if you just want to run something offline and argue with it or whatever, but variations of Llama.cpp (llamasharp in my case) + ML.net (https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fdotnet.microsoft.com%2Fen-us%2Fapps%2Fai%2Fml-dotnet) = a personal AI app (...Leroy, because reasons). All this said, I STILL haven't seen a model that can do more than spit out VERY BASIC functions (and yes I have copilot and Perplexity, and neither really impressed me), and those are usually so generic and untuned they may as well have been cut/pasted from codeproject (which, they probably were...). There's this big noise of marketing around AI that is wayyyyy overhyping actual capabilities, like by a lot folks. AI is the new IOT, which was the Cloud, which ... on and on back to the OG bubble in 97, its a hype train with Altman/Musk/Thiel as some kind of high priests selling voodoo visions of thinking machines that will solve all the worlds problems, and its all bullshit. Just like everything else today, its about money.

Records stores are doing just fine. Vinyl is back in a big way, and as someone who has a ton of it I'm thrilled about it. Vinyl purchasers are pretty eclectic/odd folks in general, so there's a pervasive mindset that supports local stores when it's possible (like the Record Store Day events, and black Friday releases). My local stores have live music and all kinds of other events to make it more of an experience, plus for the most part these guys know they are terminally dependent on the people that support them, so they tend to be pretty cool folks. The linked list was crap, none of that has come to pass as they laid it out, it's in the "well but it might also" trailing categories for almost all of their points.

Thx again for the clarifications, it really helps. We have a lot to go through, and I'm crossing my fingers that most of it will transition ok. It sounds like, from all the stuff out tonight, that most of "old" windows is still there so I'm hoping it's a straightforward process, or that its at least manageable while we make any changes. Now I just need a clean workstation to test it all out on...I wonder who's on vacation? hehe :D

Ok cool, so the next question is happens to design patterns like MVVM under this shiny new wrapper? It sounds like a native code implementation takes a lot of the roundabout out by merging the View and Model together to allow for direct access via HTML5 as opposed to XAML (for the view). If that is true, it's going to suck for folks like us who tied in viewmodels and now have to go back and put the logic into a single structure instead. Yuck, that's a lot of work.

Someone Mod this up plz, this is kind of post that keeps /. special. Thx for the corrections/clarifications Shutdown. I was wondering if they truly threw MDI/Dialogs out the window, but it sounds like it works a bit like Popup objects in WPF where multiple forms just have to live within the app boundaries (?). I'll get the tech preview off MSDN in the morning and check it out... something tells me tomorrow is going to be long and filled with tequila and tears by the end. Thanks again for the post.

I have to agree with LordLimecat's humor post here. These COULD have been significant announcements, but weren't. Instead, its a way for the MS marketing department to wallpaper the fact that the "cloud" vision of theirs is invasive and mostly useless. I don't want to have to stay connected to the internet all the time to make ANY software work, unless its specifically made to that end, such as a browser. SQL failover has been around forever by the way, so that's just pissing on me and calling it rain showing a "new and Improved" version that does the same damn thing that's been around since the 90s (yeah, I'm old, and did over 10 years with SQL Server as a DBA going back to the Sybase days, and then another 5 years as an Oracle/Solaris DBA as well). NIC binding is NOT new, and has been used forever to both provide redundant connections and to amplify bandwidth such that if one fails it's seamless to a query user. This has always been the main point of database clusters in the first place. Right now, as a commercial WPF developer, I'm just about pissed off enough to go back to zero with our current .NET product iteration and start over in Linux or IOS, just to get out from under this kind of idiocy. What really makes this whole process shitty is that they pushed everyone towards silverlight, .NET, and WPF all the way up through now, waited until the shit actually started to work right for once, then threw everyone under the bus the first time something shiny ran by (HTML5). This crap with the BUILD Visual Studio demo where some jackass "changes 2 lines of code and it runs" on stage is just more marketing fluff, and I don't believe for a second that it will work like that in the real world. Hell, they still haven't fixed VS2010 yet, and it's FULL of bugs a year after release, so what makes you think this time it's different? Like it or not, software takes a long time to develop, and when a toolware company like MS arbitrarily tosses out 20 years of other people's working code they can expect a backlash. Conversion will be painful, expensive, and it genuinely gains almost nothing in this case. What's really amazing is how much of their own IP they are cannabilizing along the way to keep their sales people happy. Finally, I'm not alone among .NET devs in getting mad about all this. The debate on silverlight has been raging all summer on the MS dev forums, and people are NOT happy with them over their HTML5/bandwagon decisions. The last time I looked, there were tens of thousands of protest postings just from the Silverlight folk, and thousands more for WPF. Ballmer can say whatever he likes, and everybody wants to have a good time at BUILD and not get all bitchy, but in the end this whole idea is a mistake. They're trying to combine tablets and desktops, and they are just different. To summarize, Limecat's post is actually pretty funny, and I think he nailed MS to a tee, and its unfair to pick on him about something MS is doing to themselves. Marketing, like this windows 8 bs-fest, is fucking this industry up badly.

Say WHAT?!?!? I don't know ANY doctors who don't have at least a CPT reference to work from. Most have one of the plethora of iphone apps for coding lookup for sale if they don't have up to date EMR software (see CCHIT certification for more about what that is). What we do is try to make interfaces that allow a doctor to type in "turtle" (nod to top level author) to get filtered list of possibilities, rather than having to memorize codes. Coding is really an antique process, so as programmers its our job to make it a plain english experience for doctors. What you are missing is that billing codes and diagnostic codes are two interrelated things though, and that the bill (while important) is really only a part of the picture. A good for instance is people with drug allergies. A good EMR system has to cross reference any pending Rx requests to make sure that something you (as a patient) are given doesn't not have an allergic affect on you. This has nothing to do with billing, but everything to do with health, and should happen in realtime, not at the front desk. To me, a good programmer sees automation of complex processes like this as opportunity (and indeed, it can be a very lucrative one, even in these trying times) instead of something to be ignored and borne. In fact... who's your doctor again? J/K :D

So much this! I'm no big fan of insurance companies, but they only care about what procedure(s) are performed and what materials were used. Observations (ie - A patient note of a "cough" when seeing a patient for med refills) end up getting wrapped up into ANONYMOUS reports, which consequently go back to places like CDC. From this, hopefully one day, we can start to track trends of things like Bird Flu before they become widespread epidemics. Right now, specifically because of consumer paranoia about privacy and insurance, there can be a lot of lag time in reporting social problems which may, in fact, affect us all. Just food for thought there.

Exactly! The whole of the coding system actually uses Vocabularies to determines the state (such as mood codes) of the patient so that the actual coding can be scoped down to represent informed use. Concepts (CUI) lead to atomic use (AUI) based on the context of the patient or subject (entity) so that by a subject symantic type you can translate the actions within a patient encounter to a VERY granular level. Consequently, regarding the top level post, it's not that people are obsessed with coding, its that its the process which drives the billing cycle in pretty much every hospital, and every insurance company, and has for over 20 years. The UMLS used to drive me nuts, until I realized that it makes more sense when you start from the HL7-RIM (Reference Info Model) first and work down from there to give the framework coherence.

Cough... SNOMED-CT - 242609008 - Bite of hamster (event) :) PS - There were actually several more observable events involving hamsters, and a total description string count of 210 involving 'Hamsters'. So, I guess you have to say "yes", its worth the money?

We do this, specifically I program and design PACS (Radiology) and EMR systems that use various coded sets. What's really interesting about the "water skis" thing people caught on NPR is that in reality there are 61 specific codes involving 'skis' alone, and that's just in English. Within the full subset of the Metathesaurus we use, there are in excess of 4M interacting concepts, and with the entire set there are over 400M interacting concepts. To make things even more complex, CPT, ICD-10, SNOMED-CT, and others all have to be made relational such that an injury related to "water skis on fire" has subrelational codes for the skis (material), what the patient was doing (observational), and what a role (doctor) does about it (procedural) before it can be billed to insurance, a process that (even electronically) takes up to 30 days and 20 revised submissions. This is PER BILLING STATEMENT. You have to be a special kind of warped to get into this (I, for instance, started as a DBA...). All this said, the whole of the codesets actually can describe pretty much anything in reality, starting with the semantics of the subject and trickling down to the price of each material and service consumed. It's pretty complex, but it's truly challenging, and beats the hell out of coding drudgery based projects by a long shot.

I think this kind of shit gives me hope that the internet isn't just for porn and poker anymore. What a great and smart way to exponentially increase the resources of the project. Stuff like this, and folding@home, and other crowd-sourced projects are an amazing phenomenon. Whoever is in charge knows how to sell it too, with the whole "read it while you translate it" concept, which kind of turns the whole thing into a video game. Nice work by the Oxford folk.

Hmm this isn't really true now is it? For the record, NOBODY knew what the actual final standard for US healthcare would be until Sept 2010 (although the preliminary final ruling was July 2010, but still...). While ICD9 to 10 may have taken ten years, there was no reason to believe it was going to float to the top of the coding systems expected for use until it was written into the federal register last year. And its not like there's a ton of options or anything that could have replaced it, making early adoption risky and potentially pointless or anything.

What I think TFA is trying to express is NOT limited to ICD. That's a "low hanging fruit" which doesn't include the other coding enhancements required for EHR conformance. ICD is a way of expressing complains and maladies via a coded system so that everyone can understand it if it appears as a part of your public medical record. In other words, everyone has to say code xxxx1 = "breast cancer" so that when you show up in an ER 6 months after a diagnosis, they can tell what's going on. The elephant in the room isn't the malady descriptions, its the coding of procedural treatment (SNOMED-CT), laboratory procedures (LOINC), RX allocations (RXNORM and the NDF), and how its billed back to insurance (CPT), and more importantly how they interconnect.

We make EMR systems targeted towards radiology, and I can say with conviction that the whole CCHIT process has thrown GE and Siemens (and their ilk) into absolute chaos. They are, today, faced with fixing their old systems to be modern/conformant, and then trying to keep them updated going forward on a MUCH more aggressive maintenance schedule than they are used to. Oh, and every month that they can't do it, their customers will see as lost revenue from govt. reimbursements. Unenviable is an understatement, it will be a financial disaster for GE if they start losing people en masse to Epic and other new players.

As someone who makes "meaningful use" based systems, I can tell you its no joke to implement. CCHIT certification alone encompasses 25k pages of standards that have to be followed to the letter and proven via testing for qualification.

In a way, it's a strange twist that the big players (GE/Siemens/Merge) lobbied to make the qualifications as hard as they currently are in order to limit new competition, and are now sinking into the pit they themselves dug. Sick, but hilarious simultaneously.

Trolling, LOL.

Every computer made has more than a TB, including the new windows tablets (which we happen to have prototypes of, since we are a development shop). And I deal with big data every day (medical images, among other things). So just to recap, I'm sitting in front of a windows tablet with a 1.5TB drive in it right now.

Oh, and it has audio in, not to mention my mixer has wireless support (granted, most people would plug their tables into a sound card, but that's not the point here).

So, again, Apple is selling last year as new, except they get to control its use?

Again, how about no.