Just in case you decide to do this again, put the foil on the outside to avoid stressing the seals around your glass with the extra reflected heat and light. Depending on how old your windows are, you could either end up losing the gas between panes or cracking the glass if there are extreme temperature differentials.
Had to do this when I was dirt poor in 2009, was a lot hotter when the window seals failed....
making things fool proof is that fools are quite ingenious little buggers.
That's my specialty! My goal with my projects has always been to lower the barrier to entry while compromising as little as possible on things that matter. I remember everything Google and Apple built before they started dumbing it down and removing features. I remember how much effort went into onboarding users and making sure people could adopt things easily. It didn't always work, but I've been very successful in making my entire de-google'd stack useful to my own family.
The hardest part is done, now I just need to make it replicable for others to spin up their own. If you're interested, I've got a repo scaffolded but I haven't published the actual stack installer yet: https://git.circuitforge.tech/Circuit-Forge/turaco/issues
Decouple people from big tech, allow people to run, fix, and maintain their own hardware instead of replacing it on an industrial scale every 6-9 months and centralizing ownership in data centers.
My entire stack runs off a 14kw solar array with a battery. Net negative. It's totally doable.
Edit: also I tested my stack against a stimulated 1k active users and couldn't get it to break a sweat, so I'm not saying everyone needs to do things this way, but your techiest uncle could host the entire family in a privacy cooperative, for instance
LLM and VLM are generally the type of model that needs a ton of resources to run. Birdnet uses an audio classification model, but there are dozens of types of models.
I'm building specialized task-focused pipelines which keep the memory footprint small for self-hosted users, and I'm working on a unified framework too make building other pipelines and applications much easier: https://docs.circuitforge.tech/
If you give me some details on what you're after I might be able to find you some projects or at least clarify keywords
Edit: an example of how efficient these can get, I have a little ESP32-driven waveshare display running an ONNX wake-word detection model which is absolutely nuts. 4mb psram and a dual-core 240mhz cpu and its only job is handling "hey Mycroft" and establishing a proper speech-to-text connection with my server downstairs
Specialized models can be made incredibly efficient. The big tech guys are trying to replace humans which is why their crap takes such huge servers to run. Building local-first/only just means you need to specialize the model and the pipeline to match what you're doing
I'm mostly vegetarian, excepting family traditional meals and the very rare celebratory indulgence, but yeah it's hard to give up eggs. I've tried a few times and just can't get the recipes to behave how I want them to. So I just refuse to buy anything but pasture raised or free range eggs, often from a guy I know down the road.
For sure on the art. I'm a graphic artist by education and I'm already working with another full-time artist on the application logos. When I make enough I'm planning to hire a proper web developer for the frontend/web design as that's onc area I do lean pretty heavy on the LLM's for assistance.
Funny you should mention the AI vs LLM labeling as I much prefer to call them LLM's or "models", but I'm trying to keep it accessible for non-technical people. I think I'll go back through and rename though.
Also most of the tools are deterministic, with the LLM's filling in gaps where there needs to be some amount of probabilistic interaction, like figuring out what you could cook given only the ingredients in your pantry, or rewriting a resume 26 million ways to satisfy the ATS filters.
If you don't want to use those features, nothing forcing you. They're useful for tracking and organizing without the LLM at all.
I'm in the process of launching a new business as a single dev, and I want to be upfront about my use of AI, but I notice as soon as I mention any LLM, most people assume what I've built is "vibe coded" without even looking at the applications themselves.
I spent almost two months just setting up the devops side of things before I even considered publishing. Feedback buttons in the apps automatically open issues on my Forgejo, push mirrors to GitHub and Codeberg, and I do weekly progress reports internally (I stopped posting to Lemmy after people felt spammed and now I just post on the site blog)
I'm just not sure how to make it easy to tell that I'm actually putting my heart and soul into building software that should help people.
I've been the partner with ADHD and CPTSD in this picture before, still working my way out of it, but I recognize the phrases she's using and the reasons. Still not an excuse. You're doing the right things, she's not putting in the effort to overcome the discomfort and guilt she's going to have to confront when she accepts everything you say is true.
The more she denies it, the harder it gets to accept and start the process of healing.
I'm truly sorry for your loss and the pain you've suffered. It's real and your feelings are valid <3
Edit: for additional context, the work never really stops. You're not a cake. There is no "ding youre done". Literally typing this on my way to therapy....
This is the nightmare scenario for any team that built their whole workflow around a cloud API. No warning, no clear reason, no real support path. just a Google form and 60 people sitting on their hands.
The uncomfortable truth is that "terms of service" at this scale is just "we can pull the rug whenever." Anthropic isn't unique here either. OpenAI, Google, all of them have the same opaque enforcement problem. It's a big part of why I've been building tools that run on local inference by default. Not because cloud is bad, but because your users shouldn't be one vague policy complaint away from a complete outage.
Local gives you continuity even when the upstream disappears.
The loop in that article is deliberate. ATS exists to externalize the cost of screening onto applicants while giving employers plausible deniability about who gets filtered. The "objective" framing is the fraud. I built a tool for this (CircuitForge-Peregrine, open-core, local-first). Not because I think the system should exist, but because I kept watching people get filtered before any human ever saw their application. A lot of neurodivergent folks especially, people who interview brilliantly but get wrecked by the performance art of keyword optimization. Resume coaches and LinkedIn premium exist for people who can afford to play the game. Peregrine is for everyone else.
Using AI to survive AI screening is genuinely stupid and I don't love it. But I'd rather the tool be free and local than have the only options be "pay someone" or "lose."
Aka eeyore
Just in case you decide to do this again, put the foil on the outside to avoid stressing the seals around your glass with the extra reflected heat and light. Depending on how old your windows are, you could either end up losing the gas between panes or cracking the glass if there are extreme temperature differentials.
Had to do this when I was dirt poor in 2009, was a lot hotter when the window seals failed....
Trying to help get us there faster with self-hosted pipelines: https://docs.circuitforge.tech/
That's my specialty! My goal with my projects has always been to lower the barrier to entry while compromising as little as possible on things that matter. I remember everything Google and Apple built before they started dumbing it down and removing features. I remember how much effort went into onboarding users and making sure people could adopt things easily. It didn't always work, but I've been very successful in making my entire de-google'd stack useful to my own family.
The hardest part is done, now I just need to make it replicable for others to spin up their own. If you're interested, I've got a repo scaffolded but I haven't published the actual stack installer yet: https://git.circuitforge.tech/Circuit-Forge/turaco/issues
Saying it'll never happen and giving up won't make that 2% get any bigger
Make the tools easy enough that the irritation of the original platform gives them enough of a push to try something easy to use.
If you want to help beta test, I'm working on an appliance-level automated installer for provisioning home cloud servers.
Got friends who need help? Help them or send them my way (or to your nearest techie)
Decouple people from big tech, allow people to run, fix, and maintain their own hardware instead of replacing it on an industrial scale every 6-9 months and centralizing ownership in data centers.
My entire stack runs off a 14kw solar array with a battery. Net negative. It's totally doable.
Edit: also I tested my stack against a stimulated 1k active users and couldn't get it to break a sweat, so I'm not saying everyone needs to do things this way, but your techiest uncle could host the entire family in a privacy cooperative, for instance
LLM and VLM are generally the type of model that needs a ton of resources to run. Birdnet uses an audio classification model, but there are dozens of types of models.
I'm building specialized task-focused pipelines which keep the memory footprint small for self-hosted users, and I'm working on a unified framework too make building other pipelines and applications much easier: https://docs.circuitforge.tech/
If you give me some details on what you're after I might be able to find you some projects or at least clarify keywords
Edit: an example of how efficient these can get, I have a little ESP32-driven waveshare display running an ONNX wake-word detection model which is absolutely nuts. 4mb psram and a dual-core 240mhz cpu and its only job is handling "hey Mycroft" and establishing a proper speech-to-text connection with my server downstairs
Specialized models can be made incredibly efficient. The big tech guys are trying to replace humans which is why their crap takes such huge servers to run. Building local-first/only just means you need to specialize the model and the pipeline to match what you're doing
I'm mostly vegetarian, excepting family traditional meals and the very rare celebratory indulgence, but yeah it's hard to give up eggs. I've tried a few times and just can't get the recipes to behave how I want them to. So I just refuse to buy anything but pasture raised or free range eggs, often from a guy I know down the road.
Anytime someone suggests a meat, just sub a meatless alt. Daring (the brand) chicken takes marinades beautifully
Egg substitutions that kinda work: chia gel, corn starch, flax meal
Those look pretty good!
Lime frosting for pastries, lime bars (like lemon bars, but lime). Salsa is usually what I end up making xD
Also if you get weird combinations of ingredients and need an idea of what to do with them, I built an app for that
I'm working on some applications that use LLM's for some features, and everything you run yourself is 100% free:
https://circutforge.tech/
Source: https://git.opensourcesolarpunk.com/
Bruh, do you need beta testers?
Much participated on the advice!
For sure on the art. I'm a graphic artist by education and I'm already working with another full-time artist on the application logos. When I make enough I'm planning to hire a proper web developer for the frontend/web design as that's onc area I do lean pretty heavy on the LLM's for assistance.
Funny you should mention the AI vs LLM labeling as I much prefer to call them LLM's or "models", but I'm trying to keep it accessible for non-technical people. I think I'll go back through and rename though.
Also most of the tools are deterministic, with the LLM's filling in gaps where there needs to be some amount of probabilistic interaction, like figuring out what you could cook given only the ingredients in your pantry, or rewriting a resume 26 million ways to satisfy the ATS filters.
If you don't want to use those features, nothing forcing you. They're useful for tracking and organizing without the LLM at all.
I'm in the process of launching a new business as a single dev, and I want to be upfront about my use of AI, but I notice as soon as I mention any LLM, most people assume what I've built is "vibe coded" without even looking at the applications themselves.
I spent almost two months just setting up the devops side of things before I even considered publishing. Feedback buttons in the apps automatically open issues on my Forgejo, push mirrors to GitHub and Codeberg, and I do weekly progress reports internally (I stopped posting to Lemmy after people felt spammed and now I just post on the site blog)
I'm just not sure how to make it easy to tell that I'm actually putting my heart and soul into building software that should help people.
If curious: https://circuitforge.tech/
I've been the partner with ADHD and CPTSD in this picture before, still working my way out of it, but I recognize the phrases she's using and the reasons. Still not an excuse. You're doing the right things, she's not putting in the effort to overcome the discomfort and guilt she's going to have to confront when she accepts everything you say is true.
The more she denies it, the harder it gets to accept and start the process of healing.
I'm truly sorry for your loss and the pain you've suffered. It's real and your feelings are valid <3
Edit: for additional context, the work never really stops. You're not a cake. There is no "ding youre done". Literally typing this on my way to therapy....
This is the nightmare scenario for any team that built their whole workflow around a cloud API. No warning, no clear reason, no real support path. just a Google form and 60 people sitting on their hands.
The uncomfortable truth is that "terms of service" at this scale is just "we can pull the rug whenever." Anthropic isn't unique here either. OpenAI, Google, all of them have the same opaque enforcement problem. It's a big part of why I've been building tools that run on local inference by default. Not because cloud is bad, but because your users shouldn't be one vague policy complaint away from a complete outage.
Local gives you continuity even when the upstream disappears.
The loop in that article is deliberate. ATS exists to externalize the cost of screening onto applicants while giving employers plausible deniability about who gets filtered. The "objective" framing is the fraud. I built a tool for this (CircuitForge-Peregrine, open-core, local-first). Not because I think the system should exist, but because I kept watching people get filtered before any human ever saw their application. A lot of neurodivergent folks especially, people who interview brilliantly but get wrecked by the performance art of keyword optimization. Resume coaches and LinkedIn premium exist for people who can afford to play the game. Peregrine is for everyone else.
Using AI to survive AI screening is genuinely stupid and I don't love it. But I'd rather the tool be free and local than have the only options be "pay someone" or "lose."