Results for “research”

Train language models with SimPO, a reference-free preference optimization method that outperforms DPO without needing a reference model.

Train and run Mamba state-space models with O(n) complexity, achieving faster inference and longer context than Transformers.

Evaluate NVIDIA Cosmos Policy on LIBERO and RoboCasa simulation environments with headless GPU evaluation and inference profiling.

Train and experiment with a minimal GPT implementation in ~300 lines of PyTorch, from character-level Shakespeare to GPT-2 scale.

Train Mixture of Experts (MoE) models using DeepSpeed or HuggingFace, covering architectures, routing, load balancing, and expert parallelism.

Run interpretability experiments on neural network internals using nnsight, with optional NDIF remote execution for massive models.

Delegates tasks to specialized AI agents running in parallel via the Box0 platform, enabling code review, security checks, research, and documentation.

Build and maintain a persistent, interlinked markdown knowledge base that compiles knowledge once and keeps it current, with cross-references and contradiction flags.

Writes cornerstone articles in Alan Hirsch's voice with SEO/GEO architecture, corpus research, and citations to drive organic discovery.

Creates fully automated AI podcasts that research, write, and narrate complete episodes, with guidance on monetization and building a podcast network.

Frontier research tracking and trend scouting specialist. Use when the user asks to talk to Fade, requests the scout, or needs arxiv monitoring, paradigm shift detection, and emerging method analysis.

Interdisciplinary synthesis and cross-domain transfer specialist. Use when the user asks to talk to Astra, requests the synthesizer, or needs cross-field insight connections and multi-modal research directions.

Identifies high-quality leads for your product or service by analyzing your business, searching for target companies, and providing actionable contact strategies. Perfect for sales, business development, and marketing professionals.

Stays current on UI design patterns, tools, and frameworks by monitoring release notes and running targeted search queries.

Use when a research proposal needs work packages, milestones, deliverables, a timeline, a risk register, and budget-justification scaffolding, while keeping scientific claims and feasibility honest and human-owned.

Provides the shared STORM methodology, artifact layout, stage-gating contract, citation hygiene, and retrieval fallback. Use when executing any /storm:* skill (generate, research, outline, write, polish). Internal knowledge; never user-invocable.

Identifies high-quality leads for your product or service by analyzing your business, searching for target companies, and providing actionable contact strategies. Perfect for sales, business development, and marketing professionals.

Search the web using AIsa Scholar Web endpoint. Returns structured web results with titles, URLs, and snippets. Use when: the user needs web search, research, source discovery, or content extraction.

聚合最近 30 天的 Reddit、X/Twitter、YouTube、TikTok、Instagram、Hacker News、Polymarket 和 web search 结果. Use when: the user needs recent multi-source research across the last 30 days.

Generates structured case briefs from judicial opinions. Use when the user provides a court opinion and needs a case brief, case summary, or distillation of a judicial decision for legal research.

Autonomous research review loop using any OpenAI-compatible LLM API. Configure via llm-chat MCP server or environment variables. Trigger with "auto review loop llm" or "llm review".

Mine, stake, and manage LITCOIN tokens on Base blockchain using the Python SDK, with options for comprehension mining or LLM-powered research mining.

Write compelling B2B case studies using the STAR framework, with research, data visualization, and distribution formats.

Provides code examples and best practices for using Pinecone, a managed vector database for production RAG, recommendation, and semantic search applications.

Run ML workloads on Modal's serverless GPU cloud platform with auto-scaling, pay-per-second pricing, and Python-native infrastructure.

Fine-tune and align language models using reinforcement learning with TRL, including SFT, DPO, PPO, GRPO, and reward model training.

Analyzes email security including phishing dissection, SPF/DKIM/DMARC authentication checks, BEC fraud patterns, and OAuth token abuse research for authorized reviews.

Runs an offline ASO audit on canonical App Store metadata, then performs keyword gap analysis via Astro MCP to surface optimization opportunities.

Fetches real-time EPA air quality data and HUD foreclosure listings through an MCP server, enabling environmental monitoring and housing research.

Analyzes target markets, ICP segments, demand patterns, competitive terrain, and strategic whitespace to support founder decisions on positioning, fundraising, and go-to-market.

AI legal document generation powered by CellCog. Contracts, NDAs, terms of service, privacy policies, freelance agreements, employment contracts, compliance reviews, legal research. Precision legal drafting from natural language prompts. Not legal advice.

Systems ML and hardware-aware optimization specialist. Use when the user asks to talk to Killjoy, requests the systems researcher, or needs compute-aware architecture tradeoffs, distributed training, and inference optimization.

Intelligent search for agents. Multi-source retrieval with confidence scoring - web, academic, and Tavily in one unified API. Use when: the user needs web search, research, source discovery, or content extraction.

Audit datasets for structure, missingness, labeling, suspicious values, duplicate identifiers, and documentation readiness. Use when a researcher asks for data QA, codebook review, sanity checks, or pre-analysis cleanup guidance.

Personal investigator / people lookup skill. Deep background research on any person using public records, court documents, property records, social media, corporate filings, and web OSINT. Use when asked to look up, investigate, research, or find information about a person — including court records, traffic infractions, property ownership, career history, social media profiles, corporate affiliations, family connections, and contact details. Also use for "find everything about [person]", "background check", "dig up info on", "what can you find on", or any people-search request.

Researches an open-ended question — options, possible solutions, prior art, trade-offs, or how something works — and produces a durable, evidence-backed, adversarially-validated report that recommends an option without committing the team to any artifact. Use when you want to research approaches, weigh options, survey prior art or the state of the art, or understand how something works before committing to a direction. Does not diagnose a bug, failure, or root cause — use investigate. Does not specify a feature — use plan-a-feature. Does not create or update a coding standard — use coding-standard. Does not compare two concrete artifacts for gaps — use gap-analysis. Does not assess an existing module's architecture — use architectural-analysis. Does not capture feedback on Han's own skills — use han-feedback.