mixermachine

The cutting edge, max size models will likely stay in the GPU space for a long time. But these models are not needed for most general requests. With a fine tuned 30B quantisized model you can serve a large portion of requests with around 32GB of RAM. Free users will likely only get these kinds of models.

At some point we will get these models in hardware and the cost per token will be minimal.

What spec of Framework Desktop do you run this on?

I'm using the Codex Business subscription (about 30€) already for multiple months. Even there they cut back on the quota. A few months back it was hard for me to reach the limit. Now it is easier.

Still, in comparison with Claude Code, the quota of Codex is a much better deal. However, they should not make it worse...

Fully agree. I went to school in Germany and many of our textbooks were free there. Sometimes you would get a textbook that is already >= 10 years and out of shape but who cares? Especially the basic knowledge does not change often. Buying all these textbooks new every year feels like a scam to me as they are then only used for one year by the pupil.

Btw when you damaged a book beyond repair, you needed to pay the full price. Only the exercise books needed to be bought freshly as they were "used up" fully after the year. Still, they were often seen as optional.

This seems quite strange to claim. Basically every city in the developed world already has power plants on the outside and a lot of wires to get the electricity in

It exists and does degrade panels but the time horizon is pretty wide. Real world data shows something like 0.5% to 0.7% degregation per year on average. At the start the degregation is higher and but it slows down with age.

So a 20 year old panel might be at around 80% in the worst case. Often they are in much better shape. This seems like a pretty good deal to me.

Especially for hot and sunny areas solar is insane. At mid day, max heat, you get the peak production and can run your AC at full throttle. That enables you to efficiently work at nice temperatures.

The article states the same solar production numbers as your comment. I agree that the headline is overly positive but the ramp up of solar can't really be denied. Change at this scale is sadly slow in this rather conservative sector.

The biggest thing is truly that solar has now reached a price tag where it just makes sense to replace other sources. You don't need to think about the environment any more to prefer it.

No I don't own them. Some people here: https://www.mydealz.de/share-deal-from-app/2753917 mention as a downside that you have to connect to the manufacturer cloud to get the live production data from the interverter. Otherwise it is truly "plug everything together and it works".

Have a look here:

https://shop-sicatron.de/products/sicatron-910w-balkonkraftw...

You still need attachment material for your balcony.

Keep in mind that this system will input all it's power into your power grid. If you don't use the power it will go directly into the public grid.

Still, really good cost saving measure.

Crazy thing is, the cost of these systems has gone down even further since then. You can get a 800w plug-in solar set with panels and an inverter for around 200€. Shipping might be 70€ or you can pick it up locally at the dealer. Add another 50-100€ for attachment material and you are good.

So for 250 to 400€ you can get a system that will break even latest after four years, likely earlier.

Nothing will come close to Opus 4.6 here. You will be able to fit a destilled 20B to 30B model on your GPU. Gpt-oss-20B is quite good in my testing locally on a Macbook Pro M2 Pro 32GB.

The bigger downside, when you compare it to Opus or any other hosted model, is the limited context. You might be able to achieve around 30k. Hosted models often have 128k or more. Opus 4.6 has 200k as its standard and 1M in api beta mode.

Regarding the $200 subscription. For Claude Code with Opus (and also Sonnet) you need that, yes.

I had ChatGPT Codex GPT5.2 high reasoning running on my side project for multiple hours the last nights. It created a server deployment for QA and PROD + client builds. It waited for the builds to complete, got the logs from Github Actions and fixed problems. Only after 4 days of this (around 2-4 hours) active coding I reached the weekly limit for the ChatGPT Plus Plan (23€). Far better value so far.

To be fully honest, it fucked up one flyway script. I have to fix this now my self :D. Will write a note in the Agent.md to never alter existing scripts. But the work otherwise was quite solid and now my server is properly deployed. If I would switch between High reasoning for Planing and Middle reasoning for coding, I would get even more usage.

Don't get me wrong, but somebody has to operate an exit node and somehow there needs to be a consensus on the protocol + routing.

If the network is only earth bound fixed wireless, the distance might be small enough that the state comes for the operator itself... This raises the cost of running this network from just money to life threat.

Getting many open source satellites up in orbit might not be feasible.

Got to say, I like the current Android versions. In the early days I flashed my Motorola Defy every second month with some cool new ROM. Always rooted and Xposed, always enabling something new.

Now I run a S23 Ultra and after two years it still does everything I need. OneUI 8.0 and Android 16. For work (app de) I also have a Pixel 7a, always with the newest Android Beta. Also works well.

Even the entry level phones work OK to pretty good now. My Samsung A16 5G (also for work) functions surprisingly well for 150€.

Fully agree. ChatGPT is often very confident and tells me that X and Y is absolutely wrong in the code. It then answers with something worse... It also does rarely say "sorry, I was wrong" when the previous output was just plain lies. You really need to verify every answer because it is so confident.

I fully switched to Gemini 3 Pro. Looking into an Opus 4.5 subscription too.

My GF on the other side prefers ChatGPT for writing tasks quite a lot (school teacher classes 1-4).

When you have enough experience and the project fits, this is the way to go. They don't pay for your time. They pay for your output and you can bill them on the output.

Not quickly but if somebody puts enough money on the table, the fabs change too. All about cost and return. Micron just axed their brand crucial (end customer RAM and SSD) because they will only sell to database centers from now on.

Crazy times.

Let's see if this demand truly holds. I'm still unsure. Currently nobody makes money with AI. There will be a correction (if I can trust my magic crystal ball ;) )

Got to say, those prices were quite cheap. I also upgraded my home server to 32GB RAM and paid something like 55€. Now we can just wait for some bubble to pop...

Sadly everything in the general direction of RAM or SSD chips is getting more expensive because a lot of production capacity is redistributed to serve AI chips and everything around.

Even lower end GPUs are getting more expensive even if they are not really useful for AI. But they still contain <some> chips and ram which is in high demand.

So yes, Apple will likely also have to pay higher priceses when they renew their contracts.

Maybe I'm wrong but doesn't SFP evolve pretty heavily here? The newest version is from --2022-- 2016. There are also quite high data-rates involved. SATA and Smart are stable for a long time. Smart has some special commands depending on manufacturer but the core set of functions always work.

I think we would all be OK with a "please don't buy list" of HDDs that are well known to cause problems. "Model X of Manufacturer Y doesn't work well. Please buy something else."

They did not opt for this. They opted for "you have to buy our own overpriced drives". TBH this is quite sad. I recommended Synology to some people before... Feels like I have to walk back on my word.

We are talking about run-of-the-mill HDDs here with SATA 3 (2005) and SMART (<2000) interface. No product is perfect but these interfaces are very well tested and billions of machines run as expected with them. The move from them was purely for money reasons.

I currently dislike and comment every video that has out dubbing enabled. I hate this approach but it is the only way to somehow make this awful feature more visible.

As soon any certification is required for the business the chapter SLA (service level agreements) comes up. Basically every business that is medium to large has to pass some kind of certification. Most of them will pay for an annual support contract because the SLA chapter forces them to have some kind of measure to recover from architecture problems.

Also great video regarding this topic: "You Need to Be Bored. Here's Why." https://www.youtube.com/watch?v=orQKfIXMiA8

I now regularly force myself to "actively" do nothing for 15 minutes and just think.

All the things I put into my brain as "todo, please remember" at some point in time are coming back during these 15 minutes.

I get quite a lot of clarity with this exercise. As soon I pick up my phone afterwards and start browsing the clarity evaporates which feels bad. So wasting time on my phone becomes less and less appealing to me.

Lets see where this leads me. I so far wasted quite a bit of time with my phone.

Some performance bottlenecks seem to be still in. I added around 200 images and 10 videos of my last vacation manually via Intent to the Immich app on Android.

The startup of the Activity is VERY slow. Images are rendered one by one in the list view (potentially in full resolution?) and the scrolling in the list is quite slow. The upload button does not keep the "uploading" state but after some time jumps back to the initial "start upload" state. Going into background or turning the screen off sometimes stops the upload. This test was done on my Samsung S23 Ultra (so CPU power should not be the issue).

Nevertheless the upload works as expected if I stay in the app and keep the screen active. Seems like this is not really the intended way of uploading things to Immich (auto upload is).

Yer that is likely a bad implementation of the automatic confirmation feature. iOS and Android both make it possible to register a receiver for very specific SMS messages with additional permissions.

...or it is just a dump data grab

UI definitions should add very little storage space. Different images and videos might add a lot.

On my device with Android 15 I can't install your app. Google enforces a minimum compielSdk now. Maybe you can upgrade it in your build :)?