tlennon-ie

Motor-aware pacing — decomposes a goal, shows each estimate padded by the profile buffer, names why.

Motor-aware pacing — decomposes a goal, shows each estimate padded by the profile buffer, names why.

Morning brief — what changed overnight, what matters today, one next-action per project.

Transcript → four-section brief (my asks, others' asks, decisions, ambiguous items with verbatim quotes). Records decisions into the cognitive graph.

/resume command — reconstruct yesterday's mental state from the cognitive graph, without inventing facts.

Reformats responses to Answer-First when sessions run long; surfaces the user's own prior intent if past their break threshold. Also activates on design-critique contexts (Figma reviews, comp reviews, prototype walk-throughs, spec reviews) so the verdict lands before the reasoning.

Surfaces prior decision evidence on repeat-validation requests; declines further re-analysis until new information appears.

Converts prose, meeting notes, or overwhelm-dumps into Mermaid diagrams (flowchart / sequence / mindmap).

Security review checklist for AI SaaS products — OWASP Top 10 plus AI-specific threats (prompt injection, jailbreaks, system-prompt extraction, indirect injection via memory or uploads). Vendor-agnostic.

How to design multi-phase conversational AI flows that feel psychologically coherent — phase definition, transition signals, persona/voice consistency, and avoiding tone-bleed between sibling products on shared infrastructure.

How to keep a fleet of sibling SaaS products consistent — what counts as "core" vs "product-specific", how to backport bug fixes and infrastructure changes cleanly, schema migration coordination, and the consolidation-vs-duplication call.

How to build one design system that powers multiple sibling products with distinct brand identities — token layering (shared + brand-specific), dark glassmorphic patterns done accessibly, and motion as brand differentiation.

How to build in-product ad operations — abstracting multiple ad platform APIs behind one interface, programmatic AI video/image creative generation, channel-level budget allocation, creative rotation on performance signals, and admin-dashboard telemetry. Vendor-agnostic.

Patterns for the backend foundation of multi-tenant AI SaaS — auth provider + JWT to DB handoff, row-level security policy patterns, payment-provider webhook idempotency, credit-based billing with auto-refund, and admin allowlists. Vendor-agnostic.

How to architect a conversational voice AI pipeline — ephemeral per-session agents with injected context, streaming STT, post-call memory/goal extraction, and audio failure handling. Vendor-agnostic.

Use when the user wants to add support for a new diffusion model or LLM training script in Bracket. Covers the Trainer protocol, search-space declaration, curated configs, presets, and tests. Lists every currently-supported preset so you know whether the model is already wired in.

Use when the user wants to create a dataset.toml for a Bracket / musubi-tuner / sd-scripts training session, or asks "how do I configure my dataset", "what does the toml file look like", "I have a folder of images, what's next". Walks through asking only the questions the toml format actually needs (paths, resolution, repeats), validates the result, and writes the file.

Use when the user asks to install, launch, or first-run Bracket; or asks "how do I get started with Bracket". Walks through install.sh / install.ps1, launches the unified server on port 8000, and verifies the React UI loads.

Use when the user wants to run a Bracket orchestration session — picking a trainer, dataset, budget, then kicking off the search. Covers both UI-driven and CLI-driven flows, including the interactive path where you collect missing inputs (paths, prompts) from the user one question at a time. Use when they say things like "run a Bracket sweep", "fine-tune Z-Image with Bracket", "search for the best LoRA config".

Use when a Bracket run is failing, hanging, producing no scores, only running for a handful of steps, or the dashboard is showing weird state. Walks through ledger inspection, log triage, the LM Studio VRAM-leak failure mode, and the most common other failure classes.

Use when working on cull, the curation engine for AI image datasets. Provides architecture invariants, the right module to touch for each kind of change, run/test commands, and guardrails (path traversal, atomic .processing rename, structured-output schema enforcement).

Use when adding, modifying, or debugging LM Studio vision workers in cull. Covers the OpenAI-compatible chat-completions call, structured-output schema requirement, model autodetect, keepalive vs idle-unload, and the unloader helper. Pairs with cull-helper for the broader architecture.

Use when modifying cull's classification JSON schema, adding/removing fields the vision worker emits, changing scoring keys, or working on the .vision.json audit record. Authoritative pointer to vision_prompt.build_response_format() — the single source of truth.