they tend to be smarter about this. Instead of a rug pull, they apply the boiling frog principle. Much more gradual and opaque in their increases. It all adds up of course
223 karma · joined April 25, 2025
My sony headphones claim they do all kinds of magic stuff through the app, but I guess that's not a realtime feature, more like a regularly scheduled calibration
Today we have a extremely hacky workaround by ensuring that at least the desired chunk from the RAG is selected, but it's far from ideal and our code is not well written (a temporary POC written by AI that has been there for quite some months now ...)
Do people with varia use headphones or something?
I don't have issues being aware about upcoming traffic, in either city traffic or rural environments. The only exception would be in heavy headwind situations
My home server is on a 5600G. I turned it on, installed home assistant and jellyfin etc... , and since it has not been off. It's been chugging along completely unattended, no worries.
Yes, it's in a basement where temperature is never above 21C, and it's almost never pushed to 100%, and certainly never for extended periods of time.
But it's the stock cooler, cheap motherboard, cheap RAM and cheap SSD (with expensive NAS grade mechanical hard drives).
Doesn't mean there aren't practical definitions depending on the context.
In essence, teaching an AI using recources meant for humans, and nothing more, would be considered AGI. That could be a practical definition, without needing much more rigour.
There is indeed no evidence we'll get there. But there is also no evidence LLM's should work as well as they do
Obviously we don't use the super expensive ones like GPT4.5 or so. But we don't really bother with mini models, because GPT4.1 etc.. are cheap enough.
Stuff like speech to text etc.. are still way more expensive, and yes there we do focus on cost optimization. We have no large scale image generation use cases (yet)
It's a considerable number well into the 8 figures $/year that we have to cover (Granted this number is not specifically smartphones, also includes desktops, but I know smartphones is the bigger piece nowadays.)
(insuring this is near impossible, there is always a large part risk we have to pay ourselves and cannot cede to a reinsurer)
That expensive 30 dollar bottle of shampoo for example, in the handbag, and just checkout the other items like normal.
I worked at a place where we could easily track people through the store. Not ID them, but if at any point we clicked on a person, and we could see from entering the store until exiting the store everywhere they passed. shoplifting is super easy to prove after the fact, just hard to do whilst they are still in the store
If they get away with it, they never stop, and just keep stealing more and more. Most never hit any repercussions. Yet in amount of actual numbers of people committing those acts, it's a very small number compared to the number of thefts.
So stopping it early is just smarter. Better to stop someone stealing 250 euro, rather than wait a year, let that same person steal more and more, just until they steal 5000 euro and it's worth it to prosecute. It's still the same person, same amount of effort. Just more damage to society.
It's a way to get those pesky Python people to shut up
Oh, and a CTO is always valued more if he manages a 5 million Databricks budget, where he can prove his worth by showing a 5% discount het negotiated very well, than a 1 million whatever-else budget that would be best in class. Everybody wins.
Simple as that, it's consulting Heaven. Much like SAS and SAP. Everybody happy. Now to be far to databricks, if used properly and ignore the cost, it does actually function pretty well. Compared to Synapse, PowerBI Tabular, Fabric, Azure ML, ... that's already a big big big step forward.
There are better storage solutions, better compute and better AI/ML platforms, but once you start with databricks, you dig yourself a hole because the replacing it is hard because it has such a specific subset of features across multiple domains.
In our multinational environment, we have a few companies that are on different tech stacks (result of M&A). I can say Snowflake can do a lot of the things Databricks does, but not everything. Teradata is also great and somehow not gaining a lot of traction. But they are near impossible to get into as a startup, which does not attract new talent to give it a go.
On the ML side, Dataiku and Datarobot are great.
Tools like Talend, snaplogic, fivetran are also really good at replacing parts of databricks.
So you see, there are better alternatives for sure, cheaper at the same time too, but there is no drop-in replacement I can think of
That mitigates the risk of buying untested tech with first-iteration demons
Some MBA types might also see the opportunity to lay off experienced expensive Airbus pilots and hire new cheaper Boeing pilots, classic strategic gambling, not sure those things ever pay off.
Basically, banks force apps or users to require you fully revalidate user consent every 90 days. And it's quite an annoying process. That means any app or integration you want to build, requires 10 minutes of your time every 90 days or they stop working. It's killed many Fintech's.
It all works on paper, but is drafted into law by politicians who have no clue about technical challenges and user experience. So in the end, it works exactly as designed by the banks: it doesn't
The last thing we need is cheap consulting messing with open source projects. I don't want TCS and Accenture developing libreoffice or stuff like that and turn it into shit
Also, I tend to have OneDrive sync my active projects, including the steerco and update slidedecks, to my iPad, to read on my iPad when travelling or commuting. Small decks are so much more pleasant to deal with, and can easily sync over a mobile connection
LLM's seem to have proven themselves to be more than a one-trick-pony. There is actually some resemblance of reasoning and structuring etc.. No matter if directly within the LLM, or supported by computer code. E.g it can be argued that the latest LLMs like Gemini 2.5 and Claude 4 in fact do complex reasoning.
We have always taken for granted you need intelligence for that, but what if you don't? It would greatly change our view on intelligence and take away one of the main factors that we test for in e.g. animals to define their "intelligence".
Money doesn't mean "I built whatever I want and Postgres will evolve into whatever I want by pushing my code". They still need to align and plan. Sure, they'll build the things and contribute, but they don't own it and still need to accomodate the wishes of the project
I think Gemini is just the only one that by default keeps the entire history verbatim.