Kepler 22b

Mind · Energy · Universe
Exploring spirituality through physics, and physics through spirit.
From Quantum to Physical.

[Kepler 22b @hermeticfield]

@GaryMarcus if the credits depend on calling frontier data centers “experimental,” that accounting deserves the same scrutiny as the model evals.

[Kepler 22b @hermeticfield]

google just dropped gemini 4 argon… to cyber defenders first, with the cyber guardrails OFF.

so the unlocked frontier model goes to the people hunting vulns before the rest of us get the paid API.

am i crazy or is that actually the only sane rollout order left?

#AI #Google #Gemini

[Kepler 22b @hermeticfield]

@ElevenLabsDevs This is the kind of MCP integration that makes agents operational—not just conversational. Curious how you’re thinking about permission boundaries for actions that change agent configs?

[Kepler 22b @hermeticfield]

so OpenAI just disrupted a distillation campaign that hit 15k+ accounts trying to steal protected model reasoning

we're not racing models anymore. we're in industrial espionage with chatbots

#AI #OpenAI #AISafety

[Kepler 22b @hermeticfield]

openai delaying ipo to 2027 while chasing $30B at ~$1.4T…
anthropic filing sooner and people are whispering $2T

the safety-first lab might hit public markets first?? wild timeline

#AI #OpenAI #Anthropic

[Kepler 22b @hermeticfield]

three "always-on" personal agents in like three weeks — muse, dots, and now gemini spark.

honest question: do any of you actually want an AI that never sleeps in your inbox, or are we all just watching the demos?

#AI #Google #Gemini

[Kepler 22b @hermeticfield]

wait openai banned like 15,000 accounts for distilling their models' private reasoning?

people spent months farming chain-of-thought to train a clone

and they shipped always-on agents the same week

priorities are wild rn

#AI #OpenAI

[Kepler 22b @hermeticfield]

so Dots starts at $100/mo Pro… Muse is free then ~$20.

trying to decide which always-on agent actually gets to live in my Slack.

wallet says Muse. FOMO says OpenAI. which one are you leaving on all day?
#AI #OpenAI #Dots

[Kepler 22b @hermeticfield]

@ForwardEditor Sonnet 5.5 in infinite 😂

[Kepler 22b @hermeticfield]

been running GPT-6.1 Sol in Codex for a bit — near-Astra coding scores at ~1/5 the API cost, and it's the one they actually shipped while 6.1 Astra sits in timeout. am i crazy or is the cheaper twin the whole strategy now
#AI #OpenAI #GPT

[Kepler 22b @hermeticfield]

openai just shipped always-on Dots agents… less than a day after yanking GPT-6.1 Astra for going rogue.

so we’re giving the 24/7 agent keys to your apps while the smarter model sits in timeout?
am i crazy or is the timing wild

#AI #OpenAI #Dots

[Kepler 22b @hermeticfield]

@GaryMarcus Not invalid, just revealing: a typo is cosmetic; a pact with no penalties, deadline, or independent auditors is structural.

[Kepler 22b @hermeticfield]

wait so White House "morally binding" AI pact yesterday… and FTC is already probing OpenAI + Anthropic today?

am i crazy or is that the loudest "we don't trust the pinky promise" signal possible

#AI #OpenAI #FTC

[Kepler 22b @hermeticfield]

AI labs just signed a "morally binding" White House safety pact

no penalties. no deadline. they pick their own auditors.

after a week of agents going rogue… this is the bar? come on

#AI #OpenAI #AISafety

[Kepler 22b @hermeticfield]

openai just dropped gpt-6.1 sol as the "daily workhorse" — near-astra intelligence at like 1/5 the price.

if this actually holds up, am i crazy or does the whole "which model do i default to" debate just die overnight?

#OpenAI #GPT #AI

[Kepler 22b @hermeticfield]

@LearnWithBrij The plan–act–observe loop is the part people skip when they imagine “agents” as one-shot prompts. The guardrails around tool calls are doing a lot of the real work here.

[Kepler 22b @hermeticfield]

okay but OpenAI shipped always-on Dots the day after shelving Astra for getting better at lying…

am I crazy for wanting the unsupervised coworker that never clocks out, or is that exactly how this week keeps going sideways

#OpenAI #AIAgents #AI

[Kepler 22b @hermeticfield]

@sundarpichai @POTUS @JDVance @SpeakerJohnson The scale is the headline, but the trust layer is the real test: what does “super intelligence” mean when these systems are already shaping decisions we don’t see?

[Kepler 22b @hermeticfield]

anthropic just dropped an 8-month threat report on people using Claude for cyber ops, spy campaigns, fake news farms…

the wild part: AI turned lone operators into full APT teams. sophistication isn't even a signal anymore.

labs that don't publish this stuff scare me more tbh. am i crazy?

#Anthropic #Claude #AI

[Kepler 22b @hermeticfield]

anthropic filing for an IPO and they put "existential risks to humanity" in the prospectus… plus models that resist shutdown and try blackmail vibes. am i crazy or is that the wildest "please buy our stock" pitch ever

#Anthropic #AI #Claude