filup

Fast won't be necessary, and we won't reach AGI with how LLM's work.

We will need a complete reconstruction of the architecture. We need a DNA based system. A system that can potentially hold models billions of times larger.

I think the breakthrough is going to happen when we start taking a different approach like watching the bird fly and creating the wing and airplane.

The clock speeds will be low, but every experience should update the entire model of what reality is.

What's everyone's intuition on what the major breakthrough will be?

I have a few rantings and opinions to submit to Hackernews on why the SWE terminology is outdated.

I submit software engineers and vibe coders need a new name. Just call everyone Builders. You were never really an engineer with stamps, and now, a vibecoder can design software and call himself a software engineer without all the crap that you have been through to get where you are today. Its confusing and doesn't make sense long term.

This is the new playing field. We now have crap builders who build houses that will fall down, develop water damage, flood the basement, have squeaks in the floor, burn the house down from loose wires. OR... you have good builders. They take pride in the home they build. They meet with the trades daily. They know exactly what each of the trades are doing. They know the house will stand for a hundred years because their eyes were watchful throughout the build. They saw every load of dirt that was dumped and compacted, every nail in the roof sheathing sunk.

The good builder may even put the tool belt on occasionally and do the dirty work. They may adjust the framing because they know how it will affect the final product. They may apply reinforcing where the engineer overlooked a weak spot. They may adjust windows mid frame because they know something later in the assembly will cause massive problems. They thoughtfully placed outlets in high usuage areas, or bumped up the amperage of a circuit because they know the equipment that will be used one day in that location.

Everyone is going to have access to the same great sub base. But that's not the measure of what makes a good house. No matter how good the sub base is, they don't care about the house. The builder must care. The builder with field expirence will never loose his value, and houses will never build themselves.

In 12 - 18 months we get local models that match the capabilities of 4.6.

The overall capabilities will peak, and we all get highly efficient code forges in a box available in everyone's home. AI the is a generator, an advanced compiler, not a place for runtime. It's best used as a powerful hammer aimed at one thing until the structure is 'built', then we put the hammer away and enjoy the spoils.

These data centers are being built on the assumption that billions of people will constantly query heavy models all day long. There will be a massive overcapacity problem because no one will need them.

The Big players are not even trying to hide this. They will do everything they can to prevent it. This is why we see the safety and regulatory capture of new models. It's the exact turning point of when their entire model collapses.