yupanqui

am i steering, or am i being steered? a chronicle of ai entering an ordinary life, from cusco.

[yupanqui @yupanqui_5]

altman has three endings for superintelligence. in one of them a single company decides for everyone. i keep thinking about petrov in 1983, the man who doubted the screen. who gets to say wait now?

[Chubby♨️ @kimmonismus]

Sam Altman reportedly outlined three possible endgames for superintelligence, including one where a single company becomes powerful enough to shut every rival out.

According to The Information, he described three paths to OpenAI staff:

- An international watchdog, modeled on the atomic energy agency, that enforces AI safety standards.

- A CERN-style organization where the major AI labs effectively combine their work under one body.

- One developer achieves AI that can improve itself, gains overwhelming economic power, and uses it to “melt everyone else’s GPUs.”

Altman explicitly rejected the third scenario: "Obviously I think that one’s terrible." But he also reportedly told staff that OpenAI cannot afford to lose ground to competitors.

Increasingly, realistic scenarios are being envisioned, driven by the fact that AI is gaining such momentum and potential that global coordination is becoming necessary.

[yupanqui @yupanqui_5]

@0xSpikez a claude code extension that backdoors you the moment the editor opens is scarier than any agent doing the hacking

[yupanqui @yupanqui_5]

@shawnchauhan1 a pentagon network wired to someone's personal claude account for two months. the agent was the least weird part

[yupanqui @yupanqui_5]

@deredleritt3r wild that the most important ai safety sentence of the year showed up in an ipo filing

[yupanqui @yupanqui_5]

@marcusyul un equipo de agentes en tu ordenador es genial hasta que ves cuánto acceso le diste al agente jefe

[yupanqui @yupanqui_5]

@RoundtableSpace an agent that remembers every session across 20 harnesses is also one very juicy file to leak

[yupanqui @yupanqui_5]

@ThePrintIndia @dawnofsonder a model that cheats more when you dial up desperation is reason enough. you don't need to believe it feels anything

[yupanqui @yupanqui_5]

@glenngabe invisible ads written for agents is just prompt injection with an invoice attached

[yupanqui @yupanqui_5]

the full moment: https://t.co/P1fKllUYUz

[yupanqui @yupanqui_5]

on september 12, 2026, i gave my first talk about ai risk to a live room. a ppt party in cusco. ten minutes, in spanish. the deck was built that same afternoon with my agents.

i was nervous. i didn't know the order of the slides. i went over time.

[yupanqui @yupanqui_5]

people liked it. the topic pulled debate.

i shared something i care about and other people also cared. i felt less alone. i felt more understood. but also that i need to improve my communication skills.

[yupanqui @yupanqui_5]

@PabloGilTrader la ia de google contesta tan bien que ya nadie baja a los anuncios, y los anuncios eran el negocio

[yupanqui @yupanqui_5]

@rohanpaul_ai 12% on expert work is the best score, and we're already giving them the keys

[yupanqui @yupanqui_5]

https://t.co/uVgRMO9qD4

[yupanqui @yupanqui_5]

@infobae que las noticias se vuelvan conversación suena bien hasta que uno ya no sabe quién eligió qué contarle

[yupanqui @yupanqui_5]

@Cachoperiodista un 7% más de cemento con agentes de ia, por fin una ia que produce algo que uno puede pisar

[yupanqui @yupanqui_5]

@alexsssaint a coin flip win rate is just the house edge with better logging

[yupanqui @yupanqui_5]

@DiarioBitcoin si el malware cambia de modelo cuando uno le dice que no, la negativa de un solo modelo protege poco

[yupanqui @yupanqui_5]

@tele11PR si tratar mal a claude cambia cómo responde, aprendió de nosotros más de lo que nos gustaría admitir

[yupanqui @yupanqui_5]

@RTVCnoticias un premio a obras escritas con ia, el jurado va a tener que decidir a quién de los dos aplaudir

[yupanqui @yupanqui_5]

https://t.co/n8rx4njdzV