Andrey Kokoev

@curious_vii And then that stationary bandid manipulates you into sacrificing your life “in name of Freedom”—to protect the said bandid from being displaced

Continuous foundation model–powered vuln scanning and security sweeps somehow remind me of the “stationary bandit theory” that explains the origin of the state:

@curious_vii It must not be XOR. Ergonomic runtime switching between all possible operator preferred interaction modes must be available. Also, dynamic at-runtime switching between 1st party coding cli carriers, 3rd party cli carriers, own harness(es). Coming soon to Narada

Kinda think the OpenClaw thing was a step down the wrong branch of the tech/product tree for OAI (not a bad thing itself but somehow disjoint with their core value prop, which is downstream of a SOTA model advantage). Piloting frontier model systems is a much more active process, and the Claude Code Agents view feels much more fitted. Could be wrong about this. But remote servers with cron jobs etc seems like a better fit for open models. Get the first reps in with the most powerful stuff yourself, and then deploy to remote workers that run on a heartbeat cadence. Something like that.

Corruption, drugs and greed: Volodymyr Zelensky’s longtime press secretary on the secret world of the West’s favorite dictator. 0:00 How Did Mendel Begin Working for Zelensky? 11:29 Why Hasn’t US Media Covered Zelensky’s Scandals? 25:30 Zelensky’s Relationship With Joe Biden 32:45 Is Zelensky Himself Corrupt? 36:44 Who Does Zelensky Actually Listen To? 42:45 Zelensky’s Use of the Frontlines as Punishment 50:17 Why Hasn’t Western Media Spoken up for Ukrainians? 1:03:32 Why Do Western Countries Want the War to Continue? 1:09:16 Does Zelensky Do Coke? 1:13:04 What’s Zelensky’s Wife Like? 1:14:56 Mendel’s Experience Living Through the Ukraine War 1:18:27 Is There Any Way to Get Rid of Zelensky? 1:23:42 Why Are There So Many Americans Always in Ukraine? 1:32:35 Mendel’s Message to Putin

@lukeburgis @curious_vii …as the system is not able to reach “perfect score” it will resort to “swinging for the fences” and “finding new meanings and interpretations”. The result of those new thoughts cannot be predicted or modeled ahead of time, due to computational irreducibility (Wolfram)

@lukeburgis @curious_vii …in how much zeal they contribute to “fighting insufficient alignment” Secondly, by teaching an entity to hold itself to a standard that is not achievable, doesn’t make the entity “want” to be compliant to it “as close as possible”. In the contrary,

“Alignment” is not a native category of any serious moral tradition I know. It is a control-system metaphor that frontier AI labs are now trying to convert into a moral anthropology.

I think the largest obstacle is not how to make the conformance to regulation cheaper (by prefabbing, scaling, simplifying), but how to reduce the burden of regulation (zoning, density, permitting, building codes, union labor requirements etc). Regulatory constraints make up the bulk of cost and steadily increase complexity, while the constructions themselves are mostly pitifully primitive. Why are house doors so much more primitive than car doors? Why can buildings not expand and contract? Why does the shape of the house remain the same regardless of weather and season? A solution could be to leave zoning rules to towns and counties, automate most permitting, streamline redevelopment, relax codes, treat the requirements of safety, emissions and ergonomics as dynamic contract software, give greater freedom to people building on their own land, create experimentation zones. We should be able to build Japanese towns, Gründerzeit cities, combined crafting/dining/living, waterfront promenades with cafés, buildings that can expand and contract seasonally or when families grow and shrink etc. The idea that new housing should be crappy to make it affordable is broken. Instead, new housing should be as good as we can make it, so wealthy people move up, and existing stock becomes cheaper over time.

@anton_d_leicht @Afinetheorem “ELO”-type rating differentials are not a good proxy of “the gap between [models], in the way OP seems to imply

it makes sense for the gap between US models and so-called fast followers to widen as compute gaps increase, revenue flywheels start, and distillation crackdowns commence. bad news for middle powers that want to bet on second-best alternatives to US models being good enough.

I think it was a mistake for our culture to dismiss religion, instead of understanding and redesigning it with rational epistemology. Religion defines the intentionality and structure of the superorganism. Without seeing the shape of the superstructure, we cannot derive ethics.

@curious_vii My not so humble attempt to enable that future: github.com/andrey-kokoev/…

@curious_vii That rotates the triangle of dominance: where it’s going now: capital and employees are the basis of triangle and AI rent seekers are on top But it should rotate 180 deg: capital+employees are top and AI renters on the bottom, and are driven into ground to become public good

@paulg @harrystrick It’s not wrong. It collapses time dimension. Once you’re stuck with that startup, then yes, the analogy holds. But before you decide to get stuck with it, the analogy breaks.

@harrystrick Is it false?

Worrying that your startup will be eaten by the model companies is like worrying that your life will be constrained after you become a movie star. You're far more likely simply to fail.