https://yesterdai.tedt.org/What changed in enterprise AI, why it matters, and what to do next. A publication of Ted Tschopp’s Public Workbench.enhttps://yesterdai.tedt.org/articles/kapitsa-local-inference-operations/https://yesterdai.tedt.org/articles/kapitsa-local-inference-operations/Sat, 24 Oct 2026 16:50:17 -0700Kapitsa v0.2.15 pairs serving fixes with qualified performance claims. Running a coding model locally makes network access, retained conversations, and service recovery decisions part of the team's work.Vendorshttps://yesterdai.tedt.org/articles/voxtral-voice-workflow/https://yesterdai.tedt.org/articles/voxtral-voice-workflow/Sat, 24 Oct 2026 15:46:52 -0700Mistral’s Voxtral 3 combines speech models and a voice-agent SDK. The operating challenge is preserving intent, corrections, and authorization as a conversation becomes a business action.Vendorshttps://yesterdai.tedt.org/articles/typed-decisions/https://yesterdai.tedt.org/articles/typed-decisions/Sat, 24 Oct 2026 15:46:52 -0700A typed-decision preprint explores reading a fixed set of options in one forward pass. The opportunity is narrower, cheaper decision work with explicit labels and a measured failure policy.Enterprise Opshttps://yesterdai.tedt.org/articles/scope-executable-policy/https://yesterdai.tedt.org/articles/scope-executable-policy/Sat, 24 Oct 2026 15:46:52 -0700SCOPE explores compiling natural-language policies into smaller predicates and explicit logic. The enterprise opportunity is a policy decision whose components can be tested, explained, and maintained separately.Securityhttps://yesterdai.tedt.org/articles/recoverable-agent-changes/https://yesterdai.tedt.org/articles/recoverable-agent-changes/Sat, 24 Oct 2026 15:46:52 -0700A new preprint proposes measuring both verifier strength and recoverability before allowing an agent to improve itself unattended. An apparently successful approval loop is not enough evidence.Enterprise Opshttps://yesterdai.tedt.org/articles/prompt-judge-evaluation/https://yesterdai.tedt.org/articles/prompt-judge-evaluation/Sat, 24 Oct 2026 15:46:52 -0700Two new studies show how wording, tool interfaces, judge choice, and language affect model comparisons. An enterprise scorecard needs to preserve those conditions and expose important reversals.Enterprise Opshttps://yesterdai.tedt.org/articles/pew-task-level-adoption/https://yesterdai.tedt.org/articles/pew-task-level-adoption/Sat, 24 Oct 2026 15:46:52 -0700Pew’s election-information survey distinguishes broad chatbot use from use for one particular purpose. Enterprise adoption measures should make the same distinction between access, activity, and useful outcomes.Policyhttps://yesterdai.tedt.org/articles/multiagent-review-tradeoffs/https://yesterdai.tedt.org/articles/multiagent-review-tradeoffs/Sat, 24 Oct 2026 15:46:52 -0700A controlled financial-document review study finds tradeoffs when work is divided among agents. Enterprises should measure misses, false alarms, and total review effort before adopting a team architecture.Enterprise Opshttps://yesterdai.tedt.org/articles/memory-needs-relevance/https://yesterdai.tedt.org/articles/memory-needs-relevance/Sat, 24 Oct 2026 15:46:52 -0700MemDev-Bench examines memory across real repository histories. Its findings make applicability and preserved process information more useful design questions than the amount of material an agent remembers.Enterprise Opshttps://yesterdai.tedt.org/articles/litellm-release-authority/https://yesterdai.tedt.org/articles/litellm-release-authority/Sat, 24 Oct 2026 15:46:52 -0700LiteLLM’s root-cause report describes a chain from exposed credentials to a malicious release. The operating question is whether apparently separate release controls depend on the same compromised authority.Securityhttps://yesterdai.tedt.org/articles/cost-aware-ticket-routing/https://yesterdai.tedt.org/articles/cost-aware-ticket-routing/Sat, 24 Oct 2026 15:46:52 -0700A cost-aware IT-routing study makes human triage an explicit option. Its results favor evaluating the whole decision policy, including the cost of a wrong route and the capacity of the review queue.Enterprise Opshttps://yesterdai.tedt.org/articles/context-sensitive-model-boundaries/https://yesterdai.tedt.org/articles/context-sensitive-model-boundaries/Sat, 24 Oct 2026 15:46:52 -0700A controlled pay-stub editing study probes how context changes model responses. It supports a focused application test, while its simulated setting limits conclusions about real-world outcomes or overall ethics.Securityhttps://yesterdai.tedt.org/articles/citation-relevance-verification/https://yesterdai.tedt.org/articles/citation-relevance-verification/Sat, 24 Oct 2026 15:46:52 -0700A biomedical claim-verification study finds that valid publication identifiers can hide irrelevant evidence. Enterprise retrieval systems should test the relationship between a claim and its source, beyond link existence.Securityhttps://yesterdai.tedt.org/articles/authorization-agent-boundaries/https://yesterdai.tedt.org/articles/authorization-agent-boundaries/Sat, 24 Oct 2026 15:46:52 -0700AuthorizationBench examines whether agents stay inside delegated scope even when they can finish the task. The enterprise decision is where permission is enforced, observed, and reviewed.Securityhttps://yesterdai.tedt.org/articles/ai-recommendation-bias/https://yesterdai.tedt.org/articles/ai-recommendation-bias/Sat, 24 Oct 2026 15:46:52 -0700Research traces how training and question wording affect AI-related recommendations. Mentioning AI more often does not prove better or worse advice; it does make evaluation of the recommendation process important.Policyhttps://yesterdai.tedt.org/articles/minimax-flash-task-economics/https://yesterdai.tedt.org/articles/minimax-flash-task-economics/Sat, 24 Oct 2026 15:21:03 -0700MiniMax’s M3.1 Flash announcement pairs lower quoted prices with agent benchmarks. The useful enterprise comparison is the cost of an accepted outcome, including retries, review, and failures.Vendorshttps://yesterdai.tedt.org/articles/meta-terms-data-governance/https://yesterdai.tedt.org/articles/meta-terms-data-governance/Sat, 24 Oct 2026 15:21:03 -0700Meta removed commercial-terms wording after scraping concerns. For enterprises, the useful response is to connect the contract review with an inventory of the content and systems the relationship touches.Policyhttps://yesterdai.tedt.org/articles/mercury-diffusion-serving/https://yesterdai.tedt.org/articles/mercury-diffusion-serving/Sat, 24 Oct 2026 15:21:03 -0700Inception’s diffusion-model release makes a case for faster interactive AI. A production evaluation still needs to measure the complete task, especially where waiting on tools or verifying actions dominates.Vendorshttps://yesterdai.tedt.org/articles/materials-structure-ml/https://yesterdai.tedt.org/articles/materials-structure-ml/Sat, 24 Oct 2026 15:21:03 -0700MIT’s materials research points to a familiar enterprise problem: useful information disappears when complex records are flattened into text. The representation deserves as much scrutiny as the model.Enterprise Opshttps://yesterdai.tedt.org/articles/llm-proxy-visibility/https://yesterdai.tedt.org/articles/llm-proxy-visibility/Sat, 24 Oct 2026 15:21:03 -0700A practitioner’s case for running an LLM proxy raises an operational question: can your team explain which application called which model, under whose credentials, and with what result?Enterprise Opshttps://yesterdai.tedt.org/articles/hidden-instructions-public-pages/https://yesterdai.tedt.org/articles/hidden-instructions-public-pages/Sat, 24 Oct 2026 15:21:03 -0700Redact’s WhatsApp investigation shows why a retrieved page can contain instructions as well as information. An enterprise assistant needs a clear rule for which material may influence its behavior.Securityhttps://yesterdai.tedt.org/articles/calibrated-disposition-drift/https://yesterdai.tedt.org/articles/calibrated-disposition-drift/Sat, 24 Oct 2026 15:21:03 -0700A new preprint examines how fine-tuning and reasoning settings can change model behavior. Its practical contribution is a reminder to validate the measurement before treating a safety score as stable.Securityhttps://yesterdai.tedt.org/articles/agent-harness-state/https://yesterdai.tedt.org/articles/agent-harness-state/Sat, 24 Oct 2026 15:21:03 -0700New research separates the cost of losing context from the cost of losing evidence that cannot be recovered. Enterprise agents need a deliberate policy for preserving task state across long runs.Enterprise Ops