Home · Research

Research

Production architecture and research synthesis for governed AI systems in regulated financial services.

PRODUCTION QUESTIONWhere should autonomy stop?ARCHITECTURE POSITIONDeterministic-first agentsRESEARCH BASISGoverned autonomyPRODUCTION QUESTIONWhat evidence proves safety?ARCHITECTURE POSITIONTrajectory assuranceRESEARCH BASISInference economicsPRODUCTION QUESTIONWhat context is defensible?ARCHITECTURE POSITIONGoverned reconstructionRESEARCH BASISMemory and drift
New series

Mind Frontier

Long-form papers on minds, machines and consciousness: a charter for comparing minds, operational definitions, construct validity, a public claim ledger, adversarial collaboration, the hard problem, access and phenomenal consciousness, the COGITATE results, what a coordinate means in a network and what spectra reveal, high-dimensional geometry, compression, dynamical systems, backpropagation, attention, the residual stream, induction heads, in-context learning, grokking, emergent abilities, and now illusionism, the Chinese Room, multiple realizability, heterophenomenology, panpsychism and Parfit on branching AI agents.

Read the series →
Complete collection

Human–Machine Intelligence Convergence

Two hundred and ten dependency-ordered essays connect mathematical foundations, learning systems, comparative philosophies of mind, consciousness science and governed human-machine architectures. Browse by question, series or semantic topic.

Enter the collection →
Fine-line convergence field joining biological and computational forms
Architecture papers

Architecture, risk and context

Deterministic-First Agents

Where should autonomy stop? A certainty-gradient architecture for placing rules, calibrated models and bounded reasoning inside a deterministic control envelope.

Model-Risk Evidence for Agentic Systems

What evidence proves safe behaviour? A trajectory-assurance model spanning evidence, tools, order, policy, recovery, approval and observed effect.

Governed Context Architecture

What context can a bank stand behind? Reconstruction, authorization, evidence coverage, temporal provenance and governed active memory.

Research synthesis

All research papers

All topics

What would it take to run a multi-agent language model system inside a global financial institution, where every action is subject to audit and every decision must be attributable to an accountable party?

Proposed architecture · schematic

Proposals cross one explicit execution boundary

Proposals cross one explicit execution boundaryA planner proposes a typed action to a policy gate. Approved actions reach a constrained executor. Policy, human review and the autonomy ledger remain outside the planner’s authority.CONTROL PLANEVersioned policyOwner · scope · expiryIndependent verificationEvidence and failure testsNamed human reviewerMaterial actions onlyEXECUTION PLANEAgent / plannerTyped action proposalPolicy gateGrant or denyConstrained executorTool allowlist + limitsproposeallowAutonomy ledgerMandate · proposal · decision · grant · tool receipt · reconciliationdecisionreadback
The planner can propose, but only the policy gate can admit execution. Record the grant and the observed effect together so that delegation remains attributable.
Position

I argue that the distance between multi-agent research demonstrations and regulated production deployment is primarily a governance gap, not a capability gap. Synthesising the 2023-2026 literature on orchestration frameworks, communication topologies, failure taxonomies and evaluation, I propose a four-part reference architecture, a certainty gradient, an autonomy ledger, topology governance and adversarial twin verification, offered as synthesis and position rather than novel empirical result.

Findings
  • Topology shape, not raw agent count, is the dominant predictor of task success, with returns diminishing beyond roughly a dozen agents (Qian et al., 2024)
  • MAST's fourteen failure modes fall into three categories, specification, inter-agent misalignment and verification, each carrying a distinct governance implication
  • Cost-controlled comparisons often narrow or eliminate the apparent advantage of complex multi-agent designs over simpler pipelines (Kapoor et al., 2024)
  • Practitioner-observed authority creep: individually reasonable delegation decisions accumulate into actions no single human reviewer would have approved, motivating the autonomy ledger
Open questions
  • Topology governance stands in direct tension with self-optimising graph research (GPTSwarm, Automated Design of Agentic Systems), which treats autonomous rewiring of the communication graph as a desirable capability
  • None of the four proposed mechanisms has been evaluated in a controlled, published sense, and a well-instrumented autonomy ledger is useless without a governance body with an explicit mandate to act on what it reveals
Design implications

A four-part reference architecture for regulated multi-agent deployment: certainty gradient, autonomy ledger, topology governance and adversarial twin verification

Read the paper

How should a long-horizon agent's memory be architected for regulated operation, given that unbounded context accumulation demonstrably degrades performance and cannot answer what is currently true?

Proposed architecture · schematic

Keep observations separate from accepted state

Keep observations separate from accepted stateCandidate facts enter a provenance gate. Accepted events are appended to the ledger; rejected facts are quarantined. A reconstruction process derives current case state for the bounded working context.WRITE PATHCandidate factSource + field + versionProvenance gateSource + field authorityEvent ledgerAppend-only historyacceptQuarantineReject reason + sourcerejectReconstructionApply precedence policyCurrent case stateReconciled facts onlyWorking contextCurrent request + stateREAD PATHPast events remainavailable for audit,outside live context.
Retrieval can surface an observation without making it true. Only source-authorised, reconciled facts enter the working state; keep the historical ledger available for audit and recovery.
Position

I argue that the dominant ever-growing buffer or memory-stream pattern conflates remembering that something happened with knowing what is currently true, and accumulates unverified content that is both an accuracy risk and a separately exploitable security vulnerability. I propose Markovian case-state reconstruction with provenance-gated writes, in which working memory is periodically rebuilt from a compact, governed, source-attributed case state rather than extended indefinitely, and every write must carry a verifiable source.

Findings
  • Larger context windows do not solve recall: mid-context information is used less reliably (Lost in the Middle), effective context falls short of advertised limits (RULER, BABILong), and accuracy degrades as irrelevant content accumulates even when the needed fact stays in place (context rot)
  • Compression, streaming and structured memory architectures address context management and recall but leave state maintenance, knowing what is currently true, implicit or unaddressed
  • Memory is an attack surface: a small number of poisoned entries can reliably backdoor an agent's future behaviour (AgentPoison), so the write path deserves at least as much security scrutiny as the read path
  • Retrospective provenance tagging of an existing agent's memory typically reveals how large a fraction has no recoverable source, often the single most persuasive evidence for migration
Open questions
  • Periodic reconstruction discards conversational narrative and adds latency and compute at each session boundary, costs that must be weighed against the context rot and unbounded accumulation it avoids
  • Provenance gating depends on a domain-specific taxonomy of acceptable sources, and deciding which business function owns that definition often takes longer than building the gate itself
Design implications

Proposes Markovian case-state reconstruction with provenance-gated writes: a governed, source-attributed case state from which the agent's working context is periodically rebuilt

Read the paper

Should the router or cascade controller that decides which model answers each request be treated as a governed model in its own right, rather than as inert infrastructure?

Proposed architecture · schematic

Cost optimisation sits inside a governed serving boundary

Cost optimisation sits inside a governed serving boundaryA request first passes cache eligibility checks. A safe hit returns an answer; a miss enters a calibrated router selecting a small, mid or frontier model. Out-of-scope requests defer, and model outputs still require acceptance checks.RequestIdentity + scopeCache eligibilityFreshness + provenanceCached answerEligible hit onlyhitCalibrated routerVersioned thresholdsmissSmall tierValidated populationMid tierValidated populationFrontier tierValidated populationALL TIERSAnswer acceptance checks → return or deferOutside calibration?Defer / human review
A cheaper route is useful only within its validated population. Preserve cache scope and freshness, a router reject option, and answer checks across every model tier.
Position

I argue that inference cost is not a single number but the product of separable decisions, which model answers, how tokens are decoded, at what precision weights are stored, and whether a cached answer suffices, and that the component coordinating these decisions is functionally a model with real financial and quality consequences. It should therefore be calibrated, drift-monitored, fallback-tested and change-controlled with the same rigour applied to the language models it selects between; the worked savings are only real if the router is governed.

Findings
  • Speculative decoding is the rare cost technique with a mathematically exact guarantee of no change to output quality, making it the easiest first step to gain model risk approval for
  • In the paper’s illustrative worked example, semantic caching, cascaded routing and quantised lower tiers reduce monthly compute cost from 300,000 to 45,816 currency units, an 84.73 percent reduction; routing is the largest lever. These are modelled values, not deployment results.
  • Cache hit rates depend on the query population, similarity threshold, freshness rule and observed error rate; benchmark rates require local validation before use in regulated customer traffic.
  • Distillation is a capital cost rather than an operating cost: it pays off for high-volume, narrow, stable tasks and is disfavoured where requirements shift frequently
Open questions
  • The routing literature implicitly assumes a fixed per-token cost per model, yet serving-system choices such as batching, paging and caching change those costs, an interaction the published research does not treat as a single governed system
  • A drifting router has no ground truth signal telling it that it has drifted, so headline savings can persist on paper while answer quality quietly degrades on unwatched traffic
Design implications

Reframes the router or cascade controller as a governed model requiring drift monitoring, a tested low-confidence fallback path, change-controlled thresholds and periodically refreshed holdout validation

Read the paper