Snippings

xxx

Most humans don’t actively optimize. They may take a good deal as they shop for a new service, but then stay with that service once the promo period ends. Banks & Fintechs count on this to justify the promo cost. It takes discipline and effort to optimize. Only a fraction make a habit of it.

Marketing departments know this and try to design propositions that deter optimization.

From: (21) Do banks need optimizer firewalls? – by Andrew M. Dresner.

xxx

xxx

Should policymakers seek to enhance the moneyness of stablecoins, we counsel for the adoption of five targeted reforms: Federal Reserve master account access for qualifying issuers, industry-funded insurance for claimholders, a secured interest regime to replace the Act’s flawed bankruptcy provisions, finality rules establishing when stablecoin transfers conclusively discharge obligations, and express tokenization of the stablecoin redemption right.

From: The Moneyness of Stablecoins by Christopher K. Odinet, Andrea Tosato, Yesha Yadav :: SSRN.

xxx

xxx

How Private Inference Works

Trusted Execution Environment (TEE)

NEAR AI Cloud combines Intel TDX and NVIDIA TEE technologies to create isolated, secure environments for AI computation:
Intel TDX(Trust Domain Extensions) : Creates confidential virtual machines (CVMs) that isolate your AI workloads from the host system, preventing unauthorized access to data in memory.
NVIDIA TEE : Provides GPU-level isolation for model inference, ensuring model weights and computations remain completely private during processing.
Cryptographic Attestation : Each TEE environment generates cryptographic proofs of its integrity and configuration, enabling independent verification of the secure execution environment.

From: Private Inference – NEAR AI.

xxx

xxx

Brave begins this journey by removing the need to trust LLM/API providers, using Confidential LLM Computing on NEAR AI TEEs. Brave uses NEAR AI TEE-enabled Nvidia GPUs to ensure confidentiality and integrity by creating secure enclaves where data and code are processed with encryption.

From: Verifiable Privacy and Transparency: A new frontier for Brave AI privacy | Brave.

xxx

xxx

When Sentinel’s decision resolves to ask the user, Sentinel creates a pending approval and execution stops. Sentinel sends that request directly to the Muse client, describing the exact action to be approved. A dialog is presented to the user directly within the client UI — not via their conversation with Muse — and their answer is routed directly back to Sentinel, which acts on the user’s decision. Sentinel then updates its authoritative approval state and allows the operation to proceed or rejects it as appropriate.

From: How We Built Safety Into Muse | Meta AI Research.

xxx

xxx

Denmark’s Digitalization Ministry said Monday that unauthorized people had accessed names, addresses and CPR numbers belonging to about 8.8 million people.

From: Stolen CPR data could make phishing attacks more convincing, cybersecurity expert warns – The Copenhagen Post.

xxx

xxx

HM Treasury and the other authorities decide what the house needs to achieve. The Retail Payments Infrastructure Board develops the plans. Its Design Authority works those plans up into something technically usable. UK PDC appoints the builders, raises the money and gets the thing built. Meanwhile, Pay.UK has to keep everybody living safely in the existing house while the new one is being constructed.

From: Meet DeliveryCo – Payments:Unpacked.

xxx

xxx

Credit Karma has actively promoted its AI-enabled Card Optimizer. I see its TV ads all the time. It optimizes consumers’ card rewards:

Centralized Point Valuations: Aggregates rewards data into one dashboard and calculates their financial value
Per-Purchase Swipe Recommendations: Alerts you on which card to use for a purchase or travel booking to maximize your rewards
Perk & Benefit Reminders: Tracks statement credits and hidden perks associated with your cards and sends alerts when benefits are expiring
Redemption Matcher: Compares ways to redeem your points to help you find the highest valuation
The service uses Open Banking to gather rewards data. It then advises consumers on how to optimize rewards accrual and rewards redemption. The consumer must consult the app first and then follow through on the recommendations manually. If they do, the card issuers are the big losers as spend economics decline and rewards reserve valuations go up (less breakage from discounted redemption deals or rewards expiration).

From: (21) Agentic Optimization emerging – by Andrew M. Dresner.

xxx

xxx

How the chain works
Each layer of the architecture relies on an open standard:

Discovery. Leverages Agent Resources Discovery (ARD) for resource and capability discovery or DNS-AID for DNS-based lookup, to find which agents exist and where they run.
Identity. Establishes a registered identity that anyone can verify from public evidence, not just a certificate the agent presents in a single session (the open Agent Name Service standard, ANS, alongside the Secure Production Identity Framework for Everyone (SPIFFE) for the identity of each running instance).
Authorization. Establishes who controls an agent and what it may do (OAuth).
Communication. Allows agents to reach tools and each other (Agent2Agent, or A2A, and the Model Context Protocol, or MCP).
Payment. Carries the person’s signed authorization and the agent’s identity into the checkout (AP2 and the Universal Commerce Protocol, or UCP).
Traceability. Records what happened and who is responsible in an append-only log, with a signed receipt that anyone can check offline (Supply Chain Integrity, Transparency and Trust).
Trust. Turns a verified identity into an evidence-based score computed from public attestations rather than self-description (the open Trust Index).

From: GoDaddy, Infoblox, Google and Cloudflare Demonstrate Secure AI Agent Transactions Across Platforms.

xxx

xxx

System-level sandboxing: The execution environment can impose additional guardrails to limit the impact of agent or model failures and facilitate detection and response. Traditional operating systems sandbox applications by statically limiting access, say to files or networks. For agents, we explore dynamically limiting agentic capabilities, establishing agent identity, and authorizing or revoking access to data based on changes in context.

From: Open and Emergent Problems in Agentic Privacy and Security: A Contextual Angle.

xxx