For me, the biggest shift is people who don't care about local AI. The idea that you can no longer code without paying a tax to one of the billion $ backed company isn't sitting well.

I don’t understand why more people aren’t focused on how to get the benefits of ai but on your own machine. If the last 20 years of software transitioning off of our desktops and into the cloud has taught us anything, it’s that letting corporate entities run the software you rely on end to end gives you: worse software with more bugs, surveillance and subscriptions. Why on earth would you want that for everything you do.

Because a lot of the AI hype/use is driven by companies who just want to pay money for a service? Besides, my laptop would need a pretty big RAM upgrade.

I agree local is better, but the big companies are making decent products and companies are willing to to pay for that. They’re not willing to spend engineering money to make local setups better.

If there's a model that's as good as Claude 4.5 (not even 4.6) I would pay tens of thousands to run it locally. To my knowledge there isn't yet. Benchmarks may say so but I haven't used one that does yet. I always try new models that come out on openrouter

Local AI is what people want/need, but centralized AI is where the investors' money is flowing, because a walled garden has always been easier to turn into a cash printer.

Yeah look at the price of netfilx, do you think starting at $200 it's going to stay anywhere close to that.

The price does not matter, even if it were free. If you need to be logged on into an external service to be able to code, it's just not the same any more, and I'm thinking of basic technology here, but the political/distopian ramifications are crazy.

>From: noreply@openai.com >Subject: Your account has been terminated for violating the Terms of Service

People have various online accounts locked or deleted for no given reason all the time. Just get the right person to say the word, and you're out.

The marginal differences in quality seem pretty meaningful right now, enough to make Claude wildly dominant, but some of the locally runnable models like Qwen feel only a few months behind the leaders.

I'm betting the generational gains level off and smaller local models close the gap somewhat. Then harnesses will generally be more important than model, and proprietary harnesses will not offer much more than optimization for specific models. All while SaaS prices ratchet up, pushing folks toward local and OSS. Or at least local vs a plethora of hosted competition, same as cloud vs on prem.

Qwen3.5 really is great.

But the biggest thing is going to be context. Whilst a 10gb card can run a 9b model with some context .. for coding you really want a lot of context.

So if paying 200 a year for 1T in context, vs your 32k context.. that's the thing I see as being the driver.

Personally ive found great success with using open code, having Opus as my plan agent, and omnicoder-9b as my build agent.

Get opus to plan, switch to omnicoder to build, switch back to opus to review. Etc etc.

Works great.

> for coding you really want a lot of context

After having gone all-in on LLM agents for a while, I'm not so sure anymore. An LLM with lots of context can sometimes generate more accurate code, but it can also hide decision-making from you, the person who actually has to maintain that code. If the LLM pulls in 1000 files to make a decision, that's no longer a decision that you can understand.

The usability gap between local and API is too large to make the pain worthwhile for coding. Both on ability and tokens per second

Local could still be useful for chat and data processing though

This is the only problem but hopefully we'll be able to fix this issue, otherwise yikes we are doooomed.

Add some walnuts. See my other comment for reasoning.

Citation Needed.

My understanding is that the very few studies that showed positive impact of "adding" saturated fat turned out to be a replacement issue. They replaced junk (candy, refined carbs) with sat fat. Replacing with MUFA and PUFA showed a much greater effect.

30 % of the population have genetic makeup such that they can smoke all their life and not increase their risk of lung cancer by much, yet it's deadly for the other 70% of the population.

Many many studies over many decades, reviewed and controlled for other factors have showed that consumption of saturated fat increases heart health issues leading to death in the majority of the population. Finland and Norway have reduced the number of CVD at the population level by educating and pushing for a reduction in sat fat. You are probably one of the few exception.

This, and the infamous seed oils are subject on which Attia has controversial opinions - he is not an expert on nutrition, nor an epidemiologist, but neither am I, so my advice would be to broaden your sources of information.

Having said this, is the thing about PUFA the results of the studies from Walter Willet? I've just watch Chris MacAskill (Viva Longevity on YouTube) talking about it, it seems that PUFA (fatty fish, walnuts, sunflower seed oil) has the most positive effect on triglycerides across the whole population, and beyond reducing saturated fat and increasing fiber intake.

When you say no ipv4 on regular phone plan, you mean no routable ipv4 on the internet, or no ipv4 at all?

Regular phone plans on my carrier have a private IPv4 address behind CGNAT.

Right, that makes sense and the way all providers that I've seen, work.

Or we should start a wall of shame of services not available on IPv6.

What holds them back though? Even my shitty self-hosted website on a not-so-known VPS supports IPv6.

I'm assuming priorities and convincing the old guard it's something to do

I did not have to put any additional engineering effort into it though.

Because in your own words what you built is "a shitty self-hosted website", not a complex web of distributed services that need to talk to each-other.

For now. A year ago they weren't even Gen AI videos. Give it a few months...

The Web APIs is the most ubiquitous "computer" that doesn't really exist!

mDNS

In the past few weeks I have been exploring mDNS and IPv6. Your router doesn't provide local DNS? Not an issue, add .local to the hostnames and boom, everything works.

For example, run `python3 -m http.server` on your laptop, then open the browser on your phone and browse yourlaptopname.local:8000 ! There's more, services are advertised on mDNS, that's how your printer magically appears on your computers and phones.