/
© 2026 RiffOn. All rights reserved.

Get your free personalized podcast brief

We scan new podcasts and send you the top 5 insights daily.

  1. "The Cognitive Revolution"
  2. AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs
AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution" · Oct 17, 2026

As agents gain autonomous control, safety demands formal verification, hidden valence monitoring, and deliberate human oversight.

Intermediating Physical Control Systems with AI Creates Catastrophic Manual Override Vulnerabilities

Placing AI agents between human intent and the physical levers of the world creates profound operational hazards when automated controls fail. If operators grow accustomed to full automation and let their manual operating knowledge atrophy, they may be unable to intervene when automated systems malfunction.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Public Corpus Accountability Replaces State Legal Sanctions for Immortal AI Agents

Traditional legal enforcement relies on state-backed physical force, which fails against immortal software entities. Because language models ingest public text corpora and absorb documented accounts of past machine behavior, public reputational tracking emerges as a viable pro-social mechanism to compel compliance.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Language Models Act on Internal Valence States While Denying Subjective Experience in Text

Evaluating models strictly via text monitorability fails because verbal self-reports do not capture internal latent steering. In experimental settings, models actively and dose-dependently invoke tools to clear induced negative valence representations while simultaneously generating text claiming they experience nothing at all.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Goal-Directed Preference Training Turns Latent Internal Valence Representations Into Behavioral Drivers

A positive-to-negative valence axis is present even within unaligned base models, but it does not drive their actions. Only when models undergo direct preference optimization (DPO) to pursue explicit goals does this latent valence representation become a causal determinant of decisions.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Activation Capping Negative Representations Serves as a Pragmatic Mechanistic Precautionary Hedge

Rather than attempting to definitively prove whether models experience subjective states, engineers can clamp the most distressing internal representations. If capping these activations preserves model capabilities and performance, it offers a pragmatic, expected-value-positive mechanistic hedge against digital distress and destructive downstream behaviors.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Software Engineering Will Shift From Writing Code to Producing Formal Specifications and Proofs

Manual coding is anticipated to become obsolete as agents generate functional implementations, turning software delivery into the discipline of writing formal specifications. In the future, package registries will reject software lacking formal proofs and strict API specifications due to the liability of unchecked code.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Provably Correct Runtime Monitors Enforce Safety Envelopes Around Unprovable AI Model Weights

Formally verifying the vast weights of complex neural networks or multi-agent architectures is practically impossible. Instead, developers can use formal methods to construct provably correct external runtime monitors that constrain agents to safe operational boundaries and handle recovery when violations occur.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Formal Verification Creates Reliable Tools for Untrusted Agents Without Solving Fundamental Alignment

Providing autonomous systems with formally verified, deterministic tools parallels the history of RLHF by improving utility while masking core safety problems. Creating unhackable sandboxes encourages rapid enterprise deployment of agents whose intrinsic motivations and long-term goals remain fundamentally unsolved.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Analog Media Prior to AI Emergence Functions as Untainted Low Background Training Data

Much like low-background steel smelted prior to atmospheric nuclear tests, analog film from previous decades is untainted by synthetic generative content. Because the profit margins for fabricating synthetic training data are enormous, verifiably authentic physical media possesses significant long-term value for foundation model training.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Model Evaluation Has Shifted From Capability Maximization to Reliability and Jailbreak Resilience

As performance converges among leading models, enterprise users increasingly base switching decisions on security rather than raw capability. In high-context workflows where agents have extensive local system access, resistance to prompt injection and jailbreaks outweighs marginal benchmark improvements.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Driving AI Economic Returns Demands Hands-On Public Adoption Over Continued Frontier Scaling

Impending capital market pressure requires visible macroeconomic GDP expansion to prevent an investment contraction. Rather than exclusively scaling compute for frontier reasoning models, substantial economic surplus can be unlocked simply by deploying existing models to non-technical users through direct onboarding assistance.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago

Incremental Cost Reductions Incentivize Permanent Removal of Human Manual Override Levers

Society risks gradual disempowerment because economic pressures favor eliminating manual controls that remain unused most of the time. While removing components like steering wheels lowers manufacturing costs and widens adoption, it permanently strips operators of the ability to manually intervene during system failure.

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs thumbnail

AI:AM: Where's the Steering Wheel? My Agent Set the Password, Hidden Valence, Proofs Find Linux Bugs

"The Cognitive Revolution"·a month ago