https://stackunseen.com What I built, tested and would use again in an enterprise: architecture maps, decisions, code and failures from real AI systems. en https://stackunseen.com/journal/build-log-unseen-ui-alpha https://stackunseen.com/journal/build-log-unseen-ui-alpha Thu, 08 Oct 2026 00:00:00 GMT How the component library behind this site went from a private workspace to two published packages, including the three things the first consumer broke. Build log unseen-uireactdeveloper-tools https://stackunseen.com/journal/weekly-note-2026-w41 https://stackunseen.com/journal/weekly-note-2026-w41 Thu, 08 Oct 2026 00:00:00 GMT The site moved to a five-area structure, every entry now hangs off a map node, and the first problem-to-production journey is live. Weekly note weeklymeta https://stackunseen.com/journal/langgraph-vs-agents-sdk https://stackunseen.com/journal/langgraph-vs-agents-sdk Wed, 07 Oct 2026 00:00:00 GMT Two ways to write the same supervisor. Compared on control flow, tracing, provider coupling, testing and what each makes hard. Comparison agentsorchestrationcomparison https://stackunseen.com/journal/weekly-note-2026-w40 https://stackunseen.com/journal/weekly-note-2026-w40 Mon, 05 Oct 2026 00:00:00 GMT Unseen UI alpha.1 reached npm, the tools page grew a radar ring, and one retrieval experiment got abandoned on purpose. Weekly note weeklyunseen-uirag https://stackunseen.com/journal/guide-production-ai-readiness https://stackunseen.com/journal/guide-production-ai-readiness Sun, 04 Oct 2026 00:00:00 GMT The last six days of a sixty-day series, and the pattern is that operability gets bought early or it does not get bought at all. Deep dive architectureobservabilitytracing https://stackunseen.com/journal/rag-production-readiness-checklist https://stackunseen.com/journal/rag-production-readiness-checklist Sat, 03 Oct 2026 00:00:00 GMT Twelve checks to pass before a retrieval system answers a real user. Tick them locally; progress stays in your browser. Checklist ragevaluationoperations https://stackunseen.com/journal/what-an-ai-gateway-costs https://stackunseen.com/journal/what-an-ai-gateway-costs Fri, 02 Oct 2026 00:00:00 GMT The operational bill for one gateway in front of a dozen tool servers, including the costs nobody budgets for. Article mcpgatewayoperationscost https://stackunseen.com/journal/working-with-claude-practices https://stackunseen.com/journal/working-with-claude-practices Wed, 30 Sep 2026 00:00:00 GMT Prompting, agents and tools, Claude Code, evaluation and safety. The habits that make Claude-based systems reliable, as a checklist you can run against your own setup. Checklist promptingagentsevaluationguardrails https://stackunseen.com/journal/outbox-pattern-for-agent-actions https://stackunseen.com/journal/outbox-pattern-for-agent-actions Mon, 28 Sep 2026 00:00:00 GMT Agents that write to systems of record need the same transactional outbox that event-driven services use. This entry covers the shape and the trade-offs. Architecture pattern agentsintegrationsarchitecturereliability https://stackunseen.com/journal/guide-agent-coordination-contracts https://stackunseen.com/journal/guide-agent-coordination-contracts Sun, 27 Sep 2026 00:00:00 GMT Six posts on multi-agent systems, and the failures were never inside an agent. They were between two of them. Deep dive multi-agentprotocolsmemory https://stackunseen.com/journal/designing-a-multi-agent-system https://stackunseen.com/journal/designing-a-multi-agent-system Sat, 26 Sep 2026 00:00:00 GMT How to design and run a multi-agent system in production, using the supervisor pattern, with its trade-offs and where human approval belongs. Architecture pattern agentsorchestrationarchitecture https://stackunseen.com/journal/when-to-bring-in-a-compliance-review https://stackunseen.com/journal/when-to-bring-in-a-compliance-review Thu, 24 Sep 2026 00:00:00 GMT The changes that should pull legal, privacy or compliance into an AI project early, and what to have ready when you do. Not legal advice; a way to ask at the right time. Checklist governancesecurityapprovals https://stackunseen.com/journal/security-review-checklist-for-ai-features https://stackunseen.com/journal/security-review-checklist-for-ai-features Tue, 22 Sep 2026 00:00:00 GMT What to check before an assistant, RAG app or agent goes in front of real users. Grouped by area, ticked off locally; progress stays in your browser. Checklist securityidentity-and-accessprompt-injectionguardrails https://stackunseen.com/journal/the-retrieval-cache-that-served-stale-policies https://stackunseen.com/journal/the-retrieval-cache-that-served-stale-policies Sun, 20 Sep 2026 00:00:00 GMT A well-meaning cache in front of retrieval kept answering from last quarter's HR policy for eleven days. Failure story ragoperationscaching https://stackunseen.com/journal/guide-multi-agent-design-patterns https://stackunseen.com/journal/guide-multi-agent-design-patterns Sat, 19 Sep 2026 00:00:00 GMT The topology gets all the design attention. Ownership, termination and traceability are what decide whether it survives contact with production. Deep dive multi-agentpipelinesprotocols https://stackunseen.com/journal/ops-drift-and-operating-cadence https://stackunseen.com/journal/ops-drift-and-operating-cadence Fri, 18 Sep 2026 00:00:00 GMT Drift is quiet by construction. SLOs and a standing review are what make it audible. Explainer driftoperations https://stackunseen.com/journal/ops-incidents-runbooks-and-rollbacks https://stackunseen.com/journal/ops-incidents-runbooks-and-rollbacks Thu, 17 Sep 2026 00:00:00 GMT AI failures cross model, prompt, retrieval, tool and data boundaries. The runbook and the rollback controls have to as well. Explainer incidentsoperations https://stackunseen.com/journal/ops-production-eval-signals https://stackunseen.com/journal/ops-production-eval-signals Wed, 16 Sep 2026 00:00:00 GMT Offline evals cover the questions you thought of. Production tells you the ones you did not. Explainer evaluationobservability https://stackunseen.com/journal/ops-cost-latency-and-capacity https://stackunseen.com/journal/ops-cost-latency-and-capacity Tue, 15 Sep 2026 00:00:00 GMT Cost and latency belong to the workflow, not the model call. Budget them before rollout, not after the invoice. Explainer costcapacity https://stackunseen.com/journal/ops-traces-logs-and-spans https://stackunseen.com/journal/ops-traces-logs-and-spans Mon, 14 Sep 2026 00:00:00 GMT A trace that links intent, prompt, retrieval, tools and output is the only thing that makes an AI failure debuggable. Explainer tracingobservability https://stackunseen.com/journal/guide-agent-control-and-supervision https://stackunseen.com/journal/guide-agent-control-and-supervision Sat, 12 Sep 2026 00:00:00 GMT Six days of notes on supervising autonomous systems, and the same failure shape kept turning up: the control exists, it is documented, and nothing in the running system is bound by it. Deep dive multi-agenthuman-reviewagents https://stackunseen.com/journal/arch-reference-architecture-and-readiness https://stackunseen.com/journal/arch-reference-architecture-and-readiness Fri, 11 Sep 2026 00:00:00 GMT Quality, safety, cost, latency, resilience and operations have to be visible on one diagram, at the same time. Article architecturereadiness https://stackunseen.com/journal/arch-approvals-events-and-tenant-boundaries https://stackunseen.com/journal/arch-approvals-events-and-tenant-boundaries Thu, 10 Sep 2026 00:00:00 GMT Approvals, events and tenant boundaries are architecture. Bolt them on as UI and they become theatre. Explainer approvalsmulti-tenancy https://stackunseen.com/journal/arch-control-plane-and-data-plane https://stackunseen.com/journal/arch-control-plane-and-data-plane Wed, 09 Sep 2026 00:00:00 GMT Policy, memory, evals and rollout get rebuilt inside every feature until someone names the layer they belong to. Explainer architecture https://stackunseen.com/journal/arch-choosing-the-right-ai-pattern https://stackunseen.com/journal/arch-choosing-the-right-ai-pattern Tue, 08 Sep 2026 00:00:00 GMT Chatbot, workflow, agent, RAG. Pick the wrong one and you spend a quarter debugging autonomy nobody asked for. Comparison architecturepatterns https://stackunseen.com/journal/arch-model-gateway-and-provider-strategy https://stackunseen.com/journal/arch-model-gateway-and-provider-strategy Mon, 07 Sep 2026 00:00:00 GMT Routing, budgets, fallbacks and policy need one place to live. Scattered SDK calls give you none of them. Explainer gatewayarchitecture https://stackunseen.com/journal/guide-agentic-ai-foundations https://stackunseen.com/journal/guide-agentic-ai-foundations Sat, 05 Sep 2026 00:00:00 GMT Six days of notes on agentic systems, and not one of the fixes that actually worked lived inside the model. Deep dive agentsguardrailsplanning https://stackunseen.com/journal/multiagent-traces-evals-and-recovery https://stackunseen.com/journal/multiagent-traces-evals-and-recovery Fri, 04 Sep 2026 00:00:00 GMT Multi-agent systems need trace timelines, trajectory evals and designed recovery, because the final output hides everything that produced it. Explainer multi-agenttracing https://stackunseen.com/journal/multiagent-permissions-budgets-stop-conditions https://stackunseen.com/journal/multiagent-permissions-budgets-stop-conditions Thu, 03 Sep 2026 00:00:00 GMT Permissions belong to roles, budgets belong to runs, and termination has to be something a machine can check. Explainer multi-agentbudgets https://stackunseen.com/journal/multiagent-who-controls-the-run https://stackunseen.com/journal/multiagent-who-controls-the-run Wed, 02 Sep 2026 00:00:00 GMT Supervisor, orchestrator-worker, planner-executor, pipeline and judge differ mainly in where control lives and when it returns. Explainer multi-agentorchestration https://stackunseen.com/journal/multiagent-role-contracts-and-handoffs https://stackunseen.com/journal/multiagent-role-contracts-and-handoffs Tue, 01 Sep 2026 00:00:00 GMT "You are a meticulous senior researcher" is a costume. Inputs, outputs, permissions and done criteria are a contract. Explainer multi-agenthandoffs https://stackunseen.com/journal/multiagent-when-multiple-agents-actually-help https://stackunseen.com/journal/multiagent-when-multiple-agents-actually-help Mon, 31 Aug 2026 00:00:00 GMT Separation of work has to buy quality, control or throughput. Usually it buys a harder debugging problem. Comparison multi-agent https://stackunseen.com/journal/guide-production-rag-operations https://stackunseen.com/journal/guide-production-rag-operations Sat, 29 Aug 2026 00:00:00 GMT Six days of notes on operating retrieval systems after launch, where nearly every real failure arrives dressed as a good answer. Deep dive ragstructured-datamultimodal https://stackunseen.com/journal/safety-data-exfiltration-and-evidence https://stackunseen.com/journal/safety-data-exfiltration-and-evidence Fri, 28 Aug 2026 00:00:00 GMT Minimise what enters context, watch the paths data can leave by, and keep the evidence an incident will demand. Explainer securityaudit https://stackunseen.com/journal/safety-blast-radius-and-containment https://stackunseen.com/journal/safety-blast-radius-and-containment Thu, 27 Aug 2026 00:00:00 GMT Narrow tools, staged writes, sandboxes and a tested kill switch. Containment is built before it is needed. Explainer securitycontainment https://stackunseen.com/journal/safety-untrusted-input-validation https://stackunseen.com/journal/safety-untrusted-input-validation Wed, 26 Aug 2026 00:00:00 GMT Injection defence is layered validation and clear authority, not a better-worded system prompt. Explainer prompt-injectionsecurity https://stackunseen.com/journal/safety-agent-identity-and-scope https://stackunseen.com/journal/safety-agent-identity-and-scope Tue, 25 Aug 2026 00:00:00 GMT Bind every AI action to a user, a service and a run, then scope the verb rather than only the data. Explainer identity-and-access https://stackunseen.com/journal/safety-architecture-not-moderation https://stackunseen.com/journal/safety-architecture-not-moderation Mon, 24 Aug 2026 00:00:00 GMT Assets, actors, abuse cases, controls, owners, evidence. Six things a team can actually write down before launch. Explainer securityguardrails https://stackunseen.com/journal/guide-evidence-driven-rag-patterns https://stackunseen.com/journal/guide-evidence-driven-rag-patterns Sat, 22 Aug 2026 00:00:00 GMT Six posts on reranking, citations and decomposition, and the uncomfortable thing they turn out to have in common. Deep dive ragretrievalreranking https://stackunseen.com/journal/evals-online-evaluation-after-launch https://stackunseen.com/journal/evals-online-evaluation-after-launch Fri, 21 Aug 2026 00:00:00 GMT Offline scores expire on contact with real users. Sampling, groundedness, drift and outcomes are the parts that keep paying. Explainer evaluation https://stackunseen.com/journal/evals-regression-gates-in-the-release-path https://stackunseen.com/journal/evals-regression-gates-in-the-release-path Thu, 20 Aug 2026 00:00:00 GMT Regression gates only work when they sit in the release path, with a threshold, a slice and an owner. Explainer evaluationrelease https://stackunseen.com/journal/evals-judges-rubrics-and-calibration https://stackunseen.com/journal/evals-judges-rubrics-and-calibration Wed, 19 Aug 2026 00:00:00 GMT Deterministic checks first, one rubric per criterion, and a calibration loop against human review. Explainer evaluation https://stackunseen.com/journal/evals-golden-set-is-a-product https://stackunseen.com/journal/evals-golden-set-is-a-product Tue, 18 Aug 2026 00:00:00 GMT Labels, edge cases, versions, an owner. Skip those and the number moves without anyone knowing why. Explainer evaluationgolden-set https://stackunseen.com/journal/evals-what-an-eval-is-allowed-to-block https://stackunseen.com/journal/evals-what-an-eval-is-allowed-to-block Mon, 17 Aug 2026 00:00:00 GMT Decide which decision the number is allowed to stop, then build the harness around that. Explainer evaluation https://stackunseen.com/journal/guide-retrieval-foundations https://stackunseen.com/journal/guide-retrieval-foundations Sat, 15 Aug 2026 00:00:00 GMT Six days of retrieval notes, and not one of the failures announced itself. Deep dive retrievalragsearch https://stackunseen.com/journal/rag-citations-evals-and-operations https://stackunseen.com/journal/rag-citations-evals-and-operations Fri, 14 Aug 2026 00:00:00 GMT Groundedness scores, per-claim citations and an operations dashboard are what tell you quality is drifting before your users do. Explainer citationsevaluation https://stackunseen.com/journal/rag-chunking-and-context-packaging https://stackunseen.com/journal/rag-chunking-and-context-packaging Thu, 13 Aug 2026 00:00:00 GMT Chunk size is a decision about what a complete answer looks like. Reranking, context budgets and refusal thresholds finish the job. Explainer chunkingrag https://stackunseen.com/journal/rag-retrieval-is-its-own-subsystem https://stackunseen.com/journal/rag-retrieval-is-its-own-subsystem Wed, 12 Aug 2026 00:00:00 GMT Most "the model is wrong" problems are recall problems. Hybrid search, query rewriting and versioned indexes are where they get fixed. Explainer retrievalrag https://stackunseen.com/journal/rag-ingestion-is-a-data-product https://stackunseen.com/journal/rag-ingestion-is-a-data-product Tue, 11 Aug 2026 00:00:00 GMT Parsing sets your quality ceiling, failures need a destination, and the update and delete paths are the half nobody tests. Explainer ingestionrag https://stackunseen.com/journal/rag-trusted-sources-and-freshness https://stackunseen.com/journal/rag-trusted-sources-and-freshness Mon, 10 Aug 2026 00:00:00 GMT Embeddings carry no access control lists. Ownership, permissions, freshness and lineage have to be designed before anything is indexed. Explainer freshnessrag https://stackunseen.com/journal/guide-from-llm-demo-to-ai-product https://stackunseen.com/journal/guide-from-llm-demo-to-ai-product Sat, 08 Aug 2026 00:00:00 GMT Six posts on the distance between a system that answers well once and a system people are willing to depend on. Deep dive raggroundinghallucination https://stackunseen.com/journal/mcp-gateway-field-notes https://stackunseen.com/journal/mcp-gateway-field-notes Thu, 06 Aug 2026 00:00:00 GMT Notes from building an MCP gateway. The protocol standardised how agents call tools, then stayed silent about credentials, context budgets and who may call what. That silence gets expensive as the servers multiply. Deep dive mcpgatewaysecurityidentity-and-access https://stackunseen.com/journal/infra-scheduled-runs-and-operations https://stackunseen.com/journal/infra-scheduled-runs-and-operations Wed, 05 Aug 2026 00:00:00 GMT Scheduled agent work fails quietly by default. Traces, release gates and a runbook are what make it fail loudly. Explainer schedulingoperations https://stackunseen.com/journal/infra-retries-and-idempotency https://stackunseen.com/journal/infra-retries-and-idempotency Tue, 04 Aug 2026 00:00:00 GMT Retries are the first thing teams switch on and the last thing they design for. Explainer retriesidempotency https://stackunseen.com/journal/infra-four-kinds-of-agent-state https://stackunseen.com/journal/infra-four-kinds-of-agent-state Mon, 03 Aug 2026 00:00:00 GMT Conversation, task, artifacts and product data end up in one store. Then a deletion request arrives and applies to all of it. Explainer statememory https://stackunseen.com/journal/guide-ai-systems-basics https://stackunseen.com/journal/guide-ai-systems-basics Sat, 01 Aug 2026 00:00:00 GMT Six days of production AI notes, and the model choice never once turned out to be the thing that broke. Deep dive modelsarchitecturegeneration https://stackunseen.com/journal/infra-tool-registry-is-access-control https://stackunseen.com/journal/infra-tool-registry-is-access-control Fri, 31 Jul 2026 00:00:00 GMT A list of tools is documentation. A registry decides who may act, as whom, and leaves proof behind. Explainer toolsidentity-and-access https://stackunseen.com/journal/infra-autonomy-is-a-runtime-decision https://stackunseen.com/journal/infra-autonomy-is-a-runtime-decision Thu, 30 Jul 2026 00:00:00 GMT Agents get more capable by default. They only get bounded on purpose. Explainer autonomyruntime https://stackunseen.com/journal/mcp-versus-direct-apis https://stackunseen.com/journal/mcp-versus-direct-apis Tue, 28 Jul 2026 00:00:00 GMT One question settles most of these arguments: will more than one AI application need this capability? Comparison mcpapis https://stackunseen.com/journal/mcp-connector-versus-agent https://stackunseen.com/journal/mcp-connector-versus-agent Mon, 27 Jul 2026 00:00:00 GMT Connector correctness and agent behaviour are separate problems. Most teams only test the first one. Explainer mcp https://stackunseen.com/journal/mcp-tools-resources-and-prompts https://stackunseen.com/journal/mcp-tools-resources-and-prompts Sat, 25 Jul 2026 00:00:00 GMT Choosing the wrong primitive hands the model authority a person was supposed to hold. Explainer mcp https://stackunseen.com/journal/mcp-where-your-server-runs https://stackunseen.com/journal/mcp-where-your-server-runs Fri, 24 Jul 2026 00:00:00 GMT Transport looks like a technical detail during the demo. It is really a choice about ownership, reachability and trust. Explainer mcpdeployment https://stackunseen.com/journal/mcp-what-the-protocol-standardises https://stackunseen.com/journal/mcp-what-the-protocol-standardises Thu, 23 Jul 2026 00:00:00 GMT It settles how capabilities are described and discovered. Every decision about whether to trust them is still yours. Explainer mcp https://stackunseen.com/journal/from-model-demos-to-mission-ready-ai-systems https://stackunseen.com/journal/from-model-demos-to-mission-ready-ai-systems Mon, 20 Jul 2026 00:00:00 GMT A practical map for moving from impressive AI demos to systems people can trust. Article architecture https://stackunseen.com/journal/why-governance-belongs-in-the-architecture https://stackunseen.com/journal/why-governance-belongs-in-the-architecture Sun, 19 Jul 2026 00:00:00 GMT Why governance belongs in the architecture, not in a document after launch. Explainer governance https://stackunseen.com/journal/why-production-ai-is-coordinated-infrastructure https://stackunseen.com/journal/why-production-ai-is-coordinated-infrastructure Sun, 19 Jul 2026 00:00:00 GMT Why production AI is the coordination of data, models, tools, controls, evaluation, and people. Article architecture https://stackunseen.com/journal/why-observability-is-not-optional-for-ai-systems https://stackunseen.com/journal/why-observability-is-not-optional-for-ai-systems Sat, 18 Jul 2026 00:00:00 GMT Why production AI needs traces across prompts, context, retrieval, tools, costs, and decisions. Explainer observabilitytracing https://stackunseen.com/journal/why-evaluating-the-model-is-not-enough https://stackunseen.com/journal/why-evaluating-the-model-is-not-enough Sat, 18 Jul 2026 00:00:00 GMT Why evaluating agents means measuring the whole workflow users experience. Explainer evaluation https://stackunseen.com/journal/why-retrieved-content-must-stay-untrusted https://stackunseen.com/journal/why-retrieved-content-must-stay-untrusted Sat, 18 Jul 2026 00:00:00 GMT Why retrieved content must stay data, not become authority over the system. Explainer prompt-injectionsecurity https://stackunseen.com/journal/why-multi-agent-systems-need-decision-rules https://stackunseen.com/journal/why-multi-agent-systems-need-decision-rules Fri, 17 Jul 2026 00:00:00 GMT Why agents need arbitration rules before disagreement reaches the final answer. Explainer governancemulti-agent https://stackunseen.com/journal/why-handoffs-need-contracts https://stackunseen.com/journal/why-handoffs-need-contracts Wed, 15 Jul 2026 00:00:00 GMT Why reliable handoffs need structured state, evidence, open questions, and ownership. Explainer handoffscontracts https://stackunseen.com/journal/why-shared-memory-becomes-a-permission-problem https://stackunseen.com/journal/why-shared-memory-becomes-a-permission-problem Tue, 14 Jul 2026 00:00:00 GMT Why shared memory becomes an access-control problem in multi-agent systems. Explainer memoryidentity-and-access https://stackunseen.com/journal/why-agents-need-a-shared-language https://stackunseen.com/journal/why-agents-need-a-shared-language Mon, 13 Jul 2026 00:00:00 GMT Why agents need shared message contracts before they can collaborate reliably. Explainer multi-agentprotocols https://stackunseen.com/journal/why-role-design-matters-more-than-agent-count https://stackunseen.com/journal/why-role-design-matters-more-than-agent-count Sun, 12 Jul 2026 00:00:00 GMT Why agent roles need clear purpose, authority, tools, memory, and evaluation. Explainer multi-agent https://stackunseen.com/journal/when-decentralized-agent-behavior-makes-sense https://stackunseen.com/journal/when-decentralized-agent-behavior-makes-sense Sat, 11 Jul 2026 00:00:00 GMT Why decentralized agent behavior needs convergence rules before it earns production trust. Explainer multi-agent https://stackunseen.com/journal/why-ai-debate-needs-a-reliable-judge https://stackunseen.com/journal/why-ai-debate-needs-a-reliable-judge Sat, 11 Jul 2026 00:00:00 GMT Why AI debate is only useful when the judge has reliable criteria. Explainer multi-agentevaluation https://stackunseen.com/journal/why-shared-state-needs-structure-and-attribution https://stackunseen.com/journal/why-shared-state-needs-structure-and-attribution Wed, 08 Jul 2026 00:00:00 GMT Why shared state needs structure, attribution, and versioning to stay debuggable. Explainer multi-agentshared-state https://stackunseen.com/journal/why-peer-agents-need-protocols-not-vibes https://stackunseen.com/journal/why-peer-agents-need-protocols-not-vibes Mon, 06 Jul 2026 00:00:00 GMT Why peer agents need communication rules, round limits, and a final decision owner. Explainer multi-agentprotocols https://stackunseen.com/journal/why-dynamic-dispatch-needs-ownership-rules https://stackunseen.com/journal/why-dynamic-dispatch-needs-ownership-rules Sun, 05 Jul 2026 00:00:00 GMT Why dynamic dispatch needs visible routing decisions and clear final ownership. Explainer multi-agent https://stackunseen.com/journal/why-pipelines-make-ai-handoffs-inspectable https://stackunseen.com/journal/why-pipelines-make-ai-handoffs-inspectable Sun, 05 Jul 2026 00:00:00 GMT Why predictable AI workflows often benefit from pipeline structure before agent autonomy. Explainer multi-agentpipelines https://stackunseen.com/journal/why-supervisor-agents-help-and-where-they-bottleneck https://stackunseen.com/journal/why-supervisor-agents-help-and-where-they-bottleneck Sat, 04 Jul 2026 00:00:00 GMT Why supervisor agents help only when routing, review, and authority are clear. Explainer multi-agent https://stackunseen.com/journal/when-hierarchy-helps-multi-agent-work-scale https://stackunseen.com/journal/when-hierarchy-helps-multi-agent-work-scale Sat, 04 Jul 2026 00:00:00 GMT Why hierarchy helps when work has real layers of authority and responsibility. Explainer multi-agent https://stackunseen.com/journal/why-one-giant-agent-is-rarely-the-cleanest-design https://stackunseen.com/journal/why-one-giant-agent-is-rarely-the-cleanest-design Fri, 03 Jul 2026 00:00:00 GMT Why multi-agent systems should reduce complexity through specialization, not add agents for novelty. Explainer multi-agent https://stackunseen.com/journal/why-autonomy-needs-budgets https://stackunseen.com/journal/why-autonomy-needs-budgets Wed, 01 Jul 2026 00:00:00 GMT Why autonomous systems need limits on cost, time, actions, retries, and risk. Explainer budgetscost https://stackunseen.com/journal/why-production-agents-need-recovery-design https://stackunseen.com/journal/why-production-agents-need-recovery-design Mon, 29 Jun 2026 00:00:00 GMT Why retries, fallbacks, checkpoints, rollback, and escalation belong in the first design. Explainer recoveryreliability https://stackunseen.com/journal/why-guardrails-should-enable-safe-action https://stackunseen.com/journal/why-guardrails-should-enable-safe-action Sun, 28 Jun 2026 00:00:00 GMT Why guardrails should shape safe progress, not just block uncertain behavior. Explainer guardrails https://stackunseen.com/journal/where-human-review-actually-belongs-in-ai-workflows https://stackunseen.com/journal/where-human-review-actually-belongs-in-ai-workflows Sat, 27 Jun 2026 00:00:00 GMT Why human review needs timing, authority, and context to be useful. Explainer human-reviewagents https://stackunseen.com/journal/why-agent-memory-is-not-one-database https://stackunseen.com/journal/why-agent-memory-is-not-one-database Sat, 27 Jun 2026 00:00:00 GMT Why useful agent memory has to be separated, scoped, and governed. Explainer memoryagents https://stackunseen.com/journal/why-reflection-should-change-the-next-action https://stackunseen.com/journal/why-reflection-should-change-the-next-action Sat, 27 Jun 2026 00:00:00 GMT Why reflection is valuable only when it decides whether to revise, retry, escalate, or stop. Explainer reflectionagents https://stackunseen.com/journal/why-tool-access-is-where-ai-becomes-operational-risk https://stackunseen.com/journal/why-tool-access-is-where-ai-becomes-operational-risk Wed, 24 Jun 2026 00:00:00 GMT Why giving AI tools means designing permissions, observability, rollback, and approval paths. Explainer toolsidentity-and-access https://stackunseen.com/journal/why-planning-reduces-wasted-ai-actions https://stackunseen.com/journal/why-planning-reduces-wasted-ai-actions Tue, 23 Jun 2026 00:00:00 GMT Why planning matters when actions have cost, risk, or dependencies. Explainer planningagents https://stackunseen.com/journal/why-agent-behavior-is-a-loop-not-a-single-prompt https://stackunseen.com/journal/why-agent-behavior-is-a-loop-not-a-single-prompt Mon, 22 Jun 2026 00:00:00 GMT Why reliable agent behavior comes from loop design, not a single clever prompt. Explainer agents https://stackunseen.com/journal/why-agents-need-outcomes-boundaries-and-stop-conditions https://stackunseen.com/journal/why-agents-need-outcomes-boundaries-and-stop-conditions Sun, 21 Jun 2026 00:00:00 GMT Why agents need clear outcomes, boundaries, budgets, and stop conditions before autonomy. Explainer agentsguardrails https://stackunseen.com/journal/why-rag-must-be-evaluated-in-parts https://stackunseen.com/journal/why-rag-must-be-evaluated-in-parts Sun, 21 Jun 2026 00:00:00 GMT Why RAG evaluation has to measure retrieval, grounding, generation, and citations separately. Explainer evaluationrag https://stackunseen.com/journal/why-weak-evidence-should-trigger-recovery-not-confidence https://stackunseen.com/journal/why-weak-evidence-should-trigger-recovery-not-confidence Sat, 20 Jun 2026 00:00:00 GMT Why weak evidence should trigger recovery, not a fluent answer with false confidence. Explainer recoveryrag https://stackunseen.com/journal/why-reflection-only-matters-when-tied-to-evidence https://stackunseen.com/journal/why-reflection-only-matters-when-tied-to-evidence Sat, 20 Jun 2026 00:00:00 GMT Why reflection only helps when it checks the answer against evidence and changes behavior. Explainer reflectionrag https://stackunseen.com/journal/why-freshness-is-part-of-correctness https://stackunseen.com/journal/why-freshness-is-part-of-correctness Fri, 19 Jun 2026 00:00:00 GMT Why freshness becomes part of correctness when the world changes faster than your index. Explainer freshnessrag https://stackunseen.com/journal/why-evidence-is-becoming-multimodal https://stackunseen.com/journal/why-evidence-is-becoming-multimodal Wed, 17 Jun 2026 00:00:00 GMT Why evidence quality changes when the source is visual, tabular, audio, or multimodal. Explainer multimodalrag https://stackunseen.com/journal/why-production-knowledge-often-lives-in-systems-not-pdfs https://stackunseen.com/journal/why-production-knowledge-often-lives-in-systems-not-pdfs Mon, 15 Jun 2026 00:00:00 GMT Why production knowledge often lives in systems of record, not only in documents. Explainer structured-datarag https://stackunseen.com/journal/why-some-answers-live-in-relationships-not-documents https://stackunseen.com/journal/why-some-answers-live-in-relationships-not-documents Sun, 14 Jun 2026 00:00:00 GMT Why some answers depend on relationships that plain document search can miss. Explainer graph-ragrag https://stackunseen.com/journal/why-one-user-question-may-need-many-searches https://stackunseen.com/journal/why-one-user-question-may-need-many-searches Sat, 13 Jun 2026 00:00:00 GMT Why one user question may need multiple searches before the evidence is good enough. Explainer retrievalrag https://stackunseen.com/journal/why-complex-ai-questions-need-decomposition https://stackunseen.com/journal/why-complex-ai-questions-need-decomposition Sat, 13 Jun 2026 00:00:00 GMT Why complex questions become more reliable when the system answers them in parts. Explainer planningrag https://stackunseen.com/journal/why-rag-needs-both-broad-context-and-precise-evidence https://stackunseen.com/journal/why-rag-needs-both-broad-context-and-precise-evidence Wed, 10 Jun 2026 00:00:00 GMT Why good RAG often needs both precise snippets and enough surrounding context. Explainer retrievalrag https://stackunseen.com/journal/why-every-important-ai-claim-needs-provenance https://stackunseen.com/journal/why-every-important-ai-claim-needs-provenance Mon, 08 Jun 2026 00:00:00 GMT Why important AI claims need receipts that users and teams can inspect. Explainer citationsprovenance https://stackunseen.com/journal/why-reranking-is-where-retrieval-becomes-useful https://stackunseen.com/journal/why-reranking-is-where-retrieval-becomes-useful Sun, 07 Jun 2026 00:00:00 GMT How reranking turns broad retrieval candidates into evidence the answer can depend on. Explainer rerankingretrieval https://stackunseen.com/journal/why-relevant-data-can-still-be-unauthorized-data https://stackunseen.com/journal/why-relevant-data-can-still-be-unauthorized-data Sun, 07 Jun 2026 00:00:00 GMT Why relevant evidence is still wrong evidence if the user was not allowed to see it. Explainer identity-and-accessretrieval https://stackunseen.com/journal/why-vector-stores-are-infrastructure-not-magic-memory https://stackunseen.com/journal/why-vector-stores-are-infrastructure-not-magic-memory Sat, 06 Jun 2026 00:00:00 GMT Why vector databases need to be operated like infrastructure, not treated like magic memory. Explainer vector-database https://stackunseen.com/journal/why-keyword-search-and-semantic-search-both-matter https://stackunseen.com/journal/why-keyword-search-and-semantic-search-both-matter Sat, 06 Jun 2026 00:00:00 GMT Why exact words and semantic meaning both matter in production retrieval. Explainer searchretrieval https://stackunseen.com/journal/why-hybrid-search-is-often-the-practical-default https://stackunseen.com/journal/why-hybrid-search-is-often-the-practical-default Sat, 06 Jun 2026 00:00:00 GMT Why combining retrieval signals is often more practical than betting on one search method. Explainer searchretrieval https://stackunseen.com/journal/why-embeddings-are-useful-and-easy-to-overtrust https://stackunseen.com/journal/why-embeddings-are-useful-and-easy-to-overtrust Fri, 05 Jun 2026 00:00:00 GMT Why semantic similarity is useful, but not the same as correctness or trust. Explainer embeddingsrag https://stackunseen.com/journal/why-chunk-boundaries-shape-answer-quality https://stackunseen.com/journal/why-chunk-boundaries-shape-answer-quality Wed, 03 Jun 2026 00:00:00 GMT How chunk boundaries decide what evidence the system can actually retrieve. Explainer chunkingrag https://stackunseen.com/journal/why-retrieval-quality-starts-before-search https://stackunseen.com/journal/why-retrieval-quality-starts-before-search Tue, 02 Jun 2026 00:00:00 GMT Why answer quality starts when knowledge enters the system, not when the user asks. Explainer ingestionrag https://stackunseen.com/journal/why-rag-is-an-evidence-design-problem-not-a-buzzword https://stackunseen.com/journal/why-rag-is-an-evidence-design-problem-not-a-buzzword Mon, 01 Jun 2026 00:00:00 GMT Why RAG is really about trusted evidence, not just letting the model search. Explainer rag https://stackunseen.com/journal/when-to-use-a-chatbot-workflow-or-agent https://stackunseen.com/journal/when-to-use-a-chatbot-workflow-or-agent Sun, 31 May 2026 00:00:00 GMT A practical way to choose between conversation, repeatable workflows, and bounded autonomy. Comparison agentsarchitecture https://stackunseen.com/journal/why-production-ai-needs-structured-outputs https://stackunseen.com/journal/why-production-ai-needs-structured-outputs Sat, 30 May 2026 00:00:00 GMT How structured output turns model text into something software can safely use. Explainer structured-output https://stackunseen.com/journal/why-a-capable-model-is-still-not-a-product https://stackunseen.com/journal/why-a-capable-model-is-still-not-a-product Sat, 30 May 2026 00:00:00 GMT Why strong model capability still needs workflow, product, and ownership design. Article product https://stackunseen.com/journal/why-confident-ai-answers-still-need-evidence https://stackunseen.com/journal/why-confident-ai-answers-still-need-evidence Wed, 27 May 2026 00:00:00 GMT Why confident language still needs evidence, boundaries, and uncertainty paths. Explainer groundinghallucination https://stackunseen.com/journal/why-creativity-settings-are-product-decisions https://stackunseen.com/journal/why-creativity-settings-are-product-decisions Mon, 25 May 2026 00:00:00 GMT How sampling settings shape the user experience, not just the writing style. Explainer samplingproduct https://stackunseen.com/journal/why-more-context-can-make-an-ai-system-worse https://stackunseen.com/journal/why-more-context-can-make-an-ai-system-worse Sun, 24 May 2026 00:00:00 GMT Why a larger context window does not remove the need for context discipline. Explainer context https://stackunseen.com/journal/why-vague-prompts-become-vague-systems https://stackunseen.com/journal/why-vague-prompts-become-vague-systems Sun, 24 May 2026 00:00:00 GMT Why prompts become operational instructions once real users depend on the system. Explainer prompting https://stackunseen.com/journal/why-llms-feel-intelligent-but-still-generate-one-step-at-a-time https://stackunseen.com/journal/why-llms-feel-intelligent-but-still-generate-one-step-at-a-time Sat, 23 May 2026 00:00:00 GMT The practical reason LLM behavior needs constraints, validation, and repeatable output design. Explainer modelsgeneration https://stackunseen.com/journal/why-tokenization-quietly-affects-cost-limits-and-reliability https://stackunseen.com/journal/why-tokenization-quietly-affects-cost-limits-and-reliability Sat, 23 May 2026 00:00:00 GMT How a low-level text detail quietly becomes a product constraint for cost, latency, and reliability. Explainer tokenscost https://stackunseen.com/journal/why-model-choice-is-rarely-the-first-production-ai-problem https://stackunseen.com/journal/why-model-choice-is-rarely-the-first-production-ai-problem Fri, 22 May 2026 00:00:00 GMT Why production AI succeeds or fails in the system around the model, not in the model choice alone. Explainer modelsarchitecture