G

Sovereign Operator

[G @updaterLabs]

human and ai are fundamentally different architectures.
- Imperfect recollection but flexible access, Human reasoning functions in infinitely wide vector spaces.
- Perfect recollection but rigid access due to storage limitations, AI reasoning narrows around a specific context.

[G @updaterLabs]

@solopribuilds this one is a certified engagement farm. You don t really believe it yourself, dude.

[G @updaterLabs]

@vineerpasam bingo. cloneable software is getting crushed so entire departments are downscaling or being eliminated and that s where engineers are getting the boot, while the serious non cloneable high reliability type needs more devs than ever.

[G @updaterLabs]

@zack_overflow wrong. models are architecturally bound to narrow reasoning around the problem s vector space. The human brain can reason in arbitrarily wide vector spaces. Doesn t matter how good AI gets at solving bounded tasks, real neurons will outperform at ideation.

[G @updaterLabs]

@toffefae @DiggTT @xoaanya "bosses" are in delusion from lack of knowledge and about to be replaced by AI governance before engineers do. The gap between real engineers and masses is exponential. Masses can now build toys and useless apps. AI engineers can build architectures that crush entire industries.

[G @updaterLabs]

@jorgemanru and it s not just that if you open your eyes you ll see platforms like github which used to be extremely reliable are full of bugs these days and that s because poorly reviewed agent code even w very strong eng teams. in the meantime, many weaker products have become unusable

[G @updaterLabs]

@jorgemanru and agents offer solutions that
- simply don t work
- with massive functional oversights
- would actually blow up production.
On a daily basis, multiple times a day.
and only people building shit apps / useless toys don t see it.

[G @updaterLabs]

@d4m1n except model outputs are non-deterministic

[G @updaterLabs]

@skylar_b_payne corporate is becoming aware of this by copilot s default GPTs running laps around Claude s PRs. in some projects even with constrained access to the codebase. Also the recent speed increase in Claude comes from trusting vibes more / lack of verification which has made it worse.

[G @updaterLabs]

@skylar_b_payne like I said except for 3d and vision / design which anthropic is overtraining for. Everything that is reasoning based, GPTs are ahead and not by a small margin.

[G @updaterLabs]

@skylar_b_payne why do you think openai is breaking the frontier in the hardest fields and anthropic is spending time overtraining models on 3d and tweaking claude code UI, all for beginners who couldn t even tell if the model were to submit production database drop ?

[G @updaterLabs]

@sipulal @anah_sahh the problem with ai slop is not the code itself. it s the much higher rate of critical oversights which combined w unlimited confidence can take production systems down. In most serious systems this can create big financial loss but I suspect you haven t seen such a system.

[G @updaterLabs]

@_ChrisCovington @skylar_b_payne can you actually read what they are building / have you seen both in a setup where they have access to each other s PRs ?

[G @updaterLabs]

I ve got hot takes

[G @updaterLabs]

@solopribuilds If I give you an AI chat app, and it never forgets, never hallucinates you ll look at the frontend and think "how the hell is this even working?". You can t give that to an agent to clone because the underlying architecture is hidden. That s the weapon.

[G @updaterLabs]

@solopribuilds When you have an established piece of software and someone can extract the functionality into a plan and give it to an agent, reduce it into a bounded task, then yes this type of software is cooked.
What the real AI engineers are building cannot be cloned like that.

[G @updaterLabs]

@Lakeybooks @VinceBlackAU @lolitadiariess No, I m just giving u real life receipts 😹 I m not grumpy or fighting in any way just broadcasting the real world.