Results for “linear-attention”

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

当感到生活失控、注意力分散，需要系统性重获专注力时

Compresses conversation transcripts into prioritized notes using an LLM observer, consolidates them when they grow, and recovers any missed sessions without a database or vector store.

Large Language Model development, training, fine-tuning, and deployment best practices.

Slow down interpretation from observation to action. Use when students or adults need to examine assumptions in conflict, dialogue, or inquiry.

Optimizes transformer attention with Flash Attention for 2-4x speedup and 10-20x memory reduction. Use when training/running transformers with long sequences (>512 tokens), encountering GPU memory issues with attention, or need faster inference. Supports PyTorch native SDPA, flash-attn library, H100 FP8, and sliding window attention.

Watches conversations continuously and compresses them into prioritized notes, consolidating and recovering missed sessions with multiple redundancy layers.

Push the LLM to reconsider, refine, and improve its recent output. Use when user asks for deeper critique or mentions a known deeper critique method, e.g. socratic, first principles, pre-mortem, red team.

Expert in creating opening lines, thumbnails, and hooks that stop the scroll. Covers curiosity gaps, pattern interrupts, emotional triggers, and platform-specific hooks. Knows how to earn attention in the first 3 seconds without resorting to clickbait. Use when "hook, opening line, scroll stopper, headline, first line, attention grab, thumbnail, clickable, " mentioned.

Documents Linear's dark-canvas marketing design system with near-black backgrounds, lavender-blue accent, four-step surface hierarchy, aggressive negative tracking on display type, and product UI screenshots as the primary visual rhythm.

Creates digital banners for social, display, hero, and email formats, applying behavioral science principles to maximize attention and conversion.

Build a library of metacognitive prompts targeting planning, monitoring, or evaluation for a specific task. Use when developing students' thinking-about-thinking during independent work.

Push the LLM to reconsider, refine, and improve its recent output. Use when user asks for deeper critique or mentions a known deeper critique method, e.g. socratic, first principles, pre-mortem, red team.

当规划内容创作策略或评估不同内容形式的长期价值时

Redesign a blocked topic sequence into an interleaved plan with mixed practice across related topics. Use when planning units, homework schedules, or revision programmes.

Push the LLM to reconsider, refine, and improve its recent output. Use when user asks for deeper critique or mentions a known deeper critique method, e.g. socratic, first principles, pre-mortem, red team.

Designs persistent memory architectures for AI agents, covering cross-session knowledge retention, entity tracking, temporal validity, graph/vector retrieval, and memory consolidation.

Design privacy-aware observability for AI agents using traces, spans, structured events, metrics, cost attribution, dashboards, alerts, and investigation workflows. Use when instrumenting an agent, debugging intermittent tool or model failures, defining service-level objectives, analyzing latency or spend, auditing agent decisions, or preparing production monitoring.

Scaling Instruction-Finetuned Language Models

LoRA: Low-Rank Adaptation of Large Language Models

Read-only OWASP LLM Top 10 audit of app-facing AI features: prompt injection, data leak, supply chain, poisoning, unsafe output, excessive agency, system-prompt leak, RAG/embedding risks, misinformation, unbounded consumption. Use when "audit LLM security", "prompt injection", "jailbreak my chatbot", "is my AI safe".

Diagnose and fix retention problems using the Fogg Behavior Model (B=MAP), covering motivation, ability, prompts, and tiny habits.

Serve LLMs and VLMs with structured outputs, prefix caching, and high throughput using RadixAttention.

当意识到自己处于迷茫、例行公事或心流等不同生活状态，并希望主动管理而非被动承受时

Apply lean thinking to UX: hypothesis-driven design, collaborative sketching, and rapid experiments instead of heavy deliverables. Use when the user mentions "Lean UX", "design hypothesis", "UX experiment", "collaborative design", "outcome over output", "design studio method", "assumption mapping", or "lightweight research". Also trigger when reducing design documentation overhead, getting cross-functional teams to co-design, or running fast usability experiments. Covers hypothesis statements, MVPs for UX, and cross-functional collaboration. For Build-Measure-Learn, see lean-startup. For usability audits, see ux-heuristics.

当观察到或经历某种极端转化现象，或试图主动促成根本性观念转变时

The layers trade timeliness against definitiveness.

Research companies, industries, or competitor sets using web search and seven analytical lenses to produce structured intelligence for downstream product management tasks.

Curating cross-agent knowledge and institutional memory: extracts patterns from agent journals into METAPATTERNS.md, detects knowledge decay, propagates best practices. Use for memory curation.

Econometrics skill for panel data models. Activates when the user asks about: "panel data", "fixed effects", "random effects", "Hausman test", "within estimator", "between estimator", "two-way fixed effects", "clustered standard errors panel", "FE model", "RE model", "pooled OLS", "unobserved heterogeneity", "panel regression", "first difference estimator", "entity fixed effects", "time fixed effects", "面板数据", "固定效应", "随机效应", "豪斯曼检验", "双向固定效应", "面板回归", "个体效应", "时间效应", "一阶差分"

LIMA: Less Is More for Alignment

Design a wise systems intervention from an existing analysis. Maps proposed actions against Meadows' leverage points, checks for unintended consequences, and generates alternatives.

Debug and optimize AI agents by analyzing reasoning traces, context degradation, tool confusion, instruction drift, repeated task failures, and performance regressions.

Econometrics skill for panel data models. Activates when the user asks about: "panel data", "fixed effects", "random effects", "Hausman test", "within estimator", "between estimator", "two-way fixed effects", "clustered standard errors panel", "FE model", "RE model", "pooled OLS", "unobserved heterogeneity", "panel regression", "first difference estimator", "entity fixed effects", "time fixed effects", "面板数据", "固定效应", "随机效应", "豪斯曼检验", "双向固定效应", "面板回归", "个体效应", "时间效应", "一阶差分"