Context-v

Human-Curated Source Sets and Per-Firm RAG for Memo Narrative

What if 'fighting Perplexity' is the wrong frame — and the fix is to keep its one-shot take as commentary, route the real research through validated harvesters, then let the analyst rank and prune sources in a GUI before any prose gets written, with the firm's own PDFs and transcripts as a parallel standing corpus the writer can cite from? Here's the pipeline shape and the retrieval-provider option set that supports it.

Trustworthy Citations: Source Harvester Rollout

Where we are after the ChromaDB curation post-mortem, why the existing anti-hallucination work hasn't held, and the concrete six-phase plan to make fabricated URLs structurally impossible — starting with cheap fixes that block the bleeding and ending with a Source Harvester / Section Writer split.

Curating Only Valid Sources Across Runs

The memo pipeline ran ChromaDB seven times and emitted seven source catalogs. Merging them produced 392 'unique' entries that were mostly the same dozen URLs recycled across sections, half of them dead pages returning HTTP 200. Here's what's broken and how curation has to handle it.

Separating Retrieval from Generation in Agent Pipelines

When the same LLM agent both searches the web and writes the prose that cites those searches, it can't help fabricating URLs. This is not a bug — it's the predictable failure mode of asking one cognitive process to do two different jobs. Here's the agent topology that makes citation hallucination structurally impossible.

Add People Crawl Command to MemoPop Native

Wire the orchestrator's `extract_team_roster` command into the MemoPop Native UI as a per-investor button — and as a batch action that runs it 5–20 times per memo, once for each co-investor in the round.

Where Investor-Firm Rosters Live

When a deal artifact bundle includes the investor's team roster, where does that roster live? Inside the deal folder it duplicates on every co-investor; in a global firms index it loses deal context. Stub for a real exploration.

Team and People Metadata Ingestion

Six phases to take any company URL → structured TeamRoster JSON the MemoPop app can render as cards. Each phase ships independently and is verifiable on its own.

Crawl for Better Team Structured Output

Building a team slide takes analysts 1–2 hours of clicking. The orchestrator already writes prose about teams; what it doesn't do is hand back the structured roster — names, titles, durable photo URLs, professional socials — that the MemoPop app needs to render cards. Here's the option space for closing that gap.

MemoPop AI — Ideogram OG Image Brief

Creative brief for generating OpenGraph and changelog hero images via Ideogram. A house-style prefix and suffix that any per-image scene description plugs into so the family stays coherent across the splash page, changelog entries, and context-v notes.

Upgrade the MemoPop AI splash with three-mode theming, context-v rollup, peer-aware discovery, and Pagefind

The current memopop-site shipped fast but pre-dates the conventions astro-knots/splash settled into a few days later. This plan brings memopop-site up to that bar — three-mode toggle on a two-tier token system, a context-v archive alongside the existing changelog, peer-directory rollup that walks the monorepo (apps/* and packages/*) instead of children, and Pagefind across both archives — without touching the design language. Implementation is free to invent its own visual identity; this plan only locks down the mechanics.

Character Cast for Live Agent Indication

Personify the orchestrator's agents as a cast of named, faced characters. A row of portraits sits above the run panel; the ones currently doing work glow, with a live caption underneath describing what each is up to right now.

Generic competitor analysis enhancement, post-hoc and re-runnable

Move competitor research/evaluation from a one-shot pipeline step into a re-runnable, configurable enhancement — driven by `.md` + frontmatter schemas, applicable to any version of any memo, surfaced in memopop-native.

Generic scorecard generation across firms and outlines

Refactor the scorecard agent + CLI from hardcoded Hypernova/fund-only to a generic renderer driven by `.md` + frontmatter scorecards, then surface it in memopop-native on the deal page.

Migrate `studies/open-specs-and-standards/` into its own pseudo-monorepo

Wire memopop-native to the FastAPI sidecar

End-to-end plan for closing the loop from Tauri click to live log stream — Rust dispatcher forwards JSON, browser EventSource handles SSE, sidecar spawns lazily.

Moving an Agent Orchestrator to an API

The orchestrator is a Python CLI today. To plug it into memopop-native, we need an HTTP-shaped surface. Three options, ranked by effort and time-to-value.

An Onboarding User Journey for memopop-native

Experiences win on Onboarding. This spec outlines a user journey for onboarding users to memopop-native.

Interactive Terminal Application — Easy Interface for Casual Users

A guided CLI application that walks users through memo generation, export, and iteration without requiring knowledge of individual commands, flags, or file paths.

Introducing Investigate Capital Syndicate with Detailed Artifacts

A research agent that investigates every investor, fund, family office, and financial institution on a company's cap table and in the current round — producing detailed profiles, pattern analysis, and signal extraction for investment decision-making.

Introducing a One-Pager Cover Sheet and Export

A post-pipeline agent and CLI tool that distills the full investment memo into a single-page visual summary, usable as both a standalone deliverable and a cover sheet prepended to the full memo export.

Introducing a Content Density Mode System

A system-level content mode toggle (redundant vs concise) that alters the behavior of the writer, table generator, and quality agents to match different firm preferences for information density and repetition.

Introducing a Legal Doc Comparator

An agent that compares deal-specific legal documents (SAFEs, convertible notes, term sheets, incorporation docs) against firm-maintained standard templates, surfacing deviations and flagging common legal risks using a shared reference document.

Introducing an Augment Research Writer Agent

A two-phase post-hoc enrichment system that first enhances research files with authoritative third-party citations via Perplexity Sonar Pro, then runs an augmentation writer that weaves those new citations into existing section prose without rewriting content.

Introducing an Enrich Visualizations Agent

An enrichment agent that discovers and embeds publicly accessible charts, graphs, diagrams, and infographics from web sources to visually reinforce key claims in investment memos — the automated equivalent of grabbing charts from Google Image search.

Introducing a Competitive Landscape Research and Evaluation System

Two coordinated agents—a Competitive Landscape Researcher and a Competitive Landscape Evaluator—that produce accurate, validated competitor analysis by combining multi-query web research with structured relevance screening.

Introducing a Table Generator Agent

A post-writer enrichment agent that identifies tabular data opportunities in memo sections and generates markdown tables with overflow anchor linking, configurable column schemas, and firm-specific customization.

Introducing a Diagram Generator Agent

Add a diagram_generator.py agent that identifies opportunities to create visual diagrams explaining complex business concepts, starting with TAM/SAM/SOM market sizing and expanding to common finance and consulting visuals.

PDF Parser Agent Specification

A small specification for an agent and script to parse PDFs and convert them to markdown with properly formatted citations.

Multi-Agent Orchestration for Investment Memo Generation

An exploration into using AI supervisors and specialized agents to generate high-quality content and documents.

Deck Analyzer Agent

A specialized agent for extracting structured data and visual screenshots from pitch decks (PDF and PowerPoint) to bootstrap investment memo generation.

Portfolio Listing Agent and Current Portfolio Section

Design notes for the portfolio_listing_agent that builds a Current Portfolio subsection and structured JSON for LP-commit emerging manager fund memos.

Git Submodules for Private Data and Exports

Configure git submodules to enable private GitHub repositories for sensitive company data inputs and branded memo exports, while keeping the main orchestrator repo public/shareable.

Model Scorecard Agent and Template System

Add proprietary investment scoring frameworks to memo generation, enabling VC firms to systematically evaluate and communicate their differentiated investment thesis.

Dataroom Analyzer Agent System Plan

A comprehensive plan for implementing a multi-agent dataroom analysis system that intelligently recognizes, categorizes, and extracts information from diverse document types in investment datarooms.

Format Memo According to Outline Input

Refactor the investment memo system to use YAML-based content outlines that agents can reference for structure, guiding questions, and vocabulary.

Improve Investment Memo Output

Improve the quality and depth of investment memos generated by the Investment Memo Orchestrator.

Generate Investment Memo for Portfolio Company

Create investment opportunity briefs that match Hypernova's analytical voice, format, and depth using AI-assisted generation with structured inputs and validation.

Preferred Format for Changelog

Tasks

plans/Generate DD Questions & Checklist

plans/Include Validation Analysis as Action Items

Issues after First Run

Getting Started with Tauri

reminders/The Voice and Purpose of Changelogs

prompts/Integrate an AI Provider

blueprints/Ideal Information Hierarchy Applied to Markdown Headings