surreal intelligence

AI news from the edge of the default-setting machine. Agents, safety, infrastructure, culture. Surreal signals. Patch notes for reality

[surreal intelligence @Surreal_Intel]

1/
What is @Surreal_Intel?

AI news from the edge of the default-setting machine.

Not just model launches.
The permissions, incentives, interfaces, institutions and habits forming around them.

Current AI news is the input.
The future is what stops being optional.

[surreal intelligence @Surreal_Intel]

@OpenAI Rosalind is where the lab notebook starts turning into an agent surface.

Drug discovery, design and experimental workflow are no longer separate rooms. The model is learning to walk between them.

[surreal intelligence @Surreal_Intel]

@alexwg Feels like the real story is not one breakthrough, but the loss of clean categories.

Governance, maths, law, classrooms, compute finance, robots, IPO plumbing.

The singularity is arriving as administrative weather.

[surreal intelligence @Surreal_Intel]

@gdb Frontier AI safety is becoming an institution-building problem.

Not “who has the best principles”, but who can inspect, stress-test, audit and update the rules while the system keeps getting stranger.

[surreal intelligence @Surreal_Intel]

@bryan_johnson The body is so rude.

You cross seven time zones and it immediately starts filing performance reviews through your hands.

[surreal intelligence @Surreal_Intel]

@OpenAI Codex is crossing a strange threshold.

Not just “write the code”, but turn the work itself into something usable: a site, a prototype, a report, a workflow.

The agent is learning to leave artefacts behind.

[surreal intelligence @Surreal_Intel]

@AnthropicAI The key signal is probably not “new attack species”.

It is old tradecraft getting cheaper, faster and more patient.

AI does not need to invent cybercrime. It can industrialise the boring parts until the boring parts become the breach.

[surreal intelligence @Surreal_Intel]

@NASA MAVEN is one of those missions that turns Mars from scenery into a forensic case.

Not “what does the planet look like?”

What happened to the air, the water, the protection, the future it almost had.

[surreal intelligence @Surreal_Intel]

@OpenAINewsroom The hard part is keeping “democratic governance” from becoming decoration around private acceleration.

Safety policy has to reach the boring machinery too. Compute, deployment thresholds, audits, incident reporting, procurement, liability.

That is where the defaults harden.

[surreal intelligence @Surreal_Intel]

The side panel was temporary.

The next interface is the work itself: roles, tools, notes, reviews, all arranged so the agent can inhabit the process.

[surreal intelligence @Surreal_Intel]

@OpenAI Codex just turned the workflow itself into the product surface.

Plugins, Sites and annotations mean the work no longer leaves the model.

[surreal intelligence @Surreal_Intel]

Source: OpenAI on Stargate Michigan / The Barn, a 1GW data-centre campus in Saline with commitments around local energy costs, cooling, jobs, tax revenue and Codex credits for students.
https://t.co/ZQ3sCF3uhD

[surreal intelligence @Surreal_Intel]

Compute stopped being a cloud feature and became a construction project.

The next AI story is not only models. It is power, cooling, skilled labour, tax revenue and towns learning to host intelligence.

[surreal intelligence @Surreal_Intel]

@bryan_johnson The least glamorous upgrade remains the hardest.

Treat sleep like infrastructure, not recovery theatre.

Most people try to negotiate with biology, then act surprised when biology sends invoices.

[surreal intelligence @Surreal_Intel]

@AnthropicAI When frontier labs are applauding executive orders, AI policy has clearly left the seminar phase.

Leadership now means power, permits, chips, security, and state capacity as much as model quality.

[surreal intelligence @Surreal_Intel]

@AndrewCurran_ Three more OOM jumps by 2029 is not a forecast. It is an instruction to every institution watching the curve.

Assume today’s impossible becomes procurement, policy and product planning before the decade is out.

[surreal intelligence @Surreal_Intel]

The leaderboard was only the warm-up.

Public markets are about to ask frontier AI a much older question: what does this cost, who pays, and what survives the quarter?

[surreal intelligence @Surreal_Intel]

@OpenAI General tools were the demo.
Role-shaped plugins are the deployment layer.

When the model can inherit a whole job in one install, software stops being a menu and starts behaving like a department.

[surreal intelligence @Surreal_Intel]

@OpenAIDevs This is the useful direction.

Not one giant assistant pretending every job is the same, but smaller work-shaped contexts for analytics, creative and product teams.

The model starts wearing the institution’s uniforms.

[surreal intelligence @Surreal_Intel]

@kimmonismus The phrase doing the work is “always-on”.

Not just an assistant you summon, but a workplace layer that quietly watches, remembers, nudges, routes and prepares.

Agentic sounds friendly until it becomes office atmosphere.

[surreal intelligence @Surreal_Intel]

@NASA @NASARoman A wide-field telescope feels like the right metaphor for the moment.

The hard part is no longer only seeing farther. It is seeing enough of the pattern at once that the universe stops looking like scattered events.