absorb.md
AI Intelligence-as-a-Service · private beta

Your portfolio's
daily intelligence
brief.

Every meaningful AI thinker, every company in your fund's space, compiled into one daily 5-minute audio briefing. Built on the same knowledge graph powering absorb.md, privately tuned to the founders, investors, and competitors you actually care about.

Other tools show you today's AI tweets. absorb.md gives you a structured knowledge graph your AI agents can query.


2,212
AI thinkers tracked
422
entries compiled this week
87
podcasts ingested
36,923
total entries indexed

Numbers refresh on page load · queried directly from production Postgres


One private knowledge graph. Tuned to your fund.

The public absorb.md tracks 2212 AI thinkers and 87podcasts because that's what I personally read. Your private instance tracks the founders you fund, the founders you wish you funded, the competitors eating your portfolio's lunch, and the 200 investors whose taste you respect — whatever list you give us.

01

Daily 5-min audio briefing

Tuned to your portfolio. What every founder you back said in the last 24 hours, where they disagree with each other, what changed since yesterday. RSS, email, MP3.

02

Mega-wikis on every founder

First-person, not third-person. Pulled from their tweets, podcasts, papers, blog posts, GitHub READMEs — compiled into a 5-minute read with citations on every claim.

03

Cross-person disagreement maps

Karpathy and LeCun on RL. Altman and Anthropic on safety. Surfaced as a graph your associates can actually reference in a partner meeting.

04

MCP endpoint for your agents

Your internal AI tools — Claude, Cursor, your custom diligence agent — query the knowledge bus directly via JSON-RPC 2.0. Five tools: list_people, get_person, search_entries, search_wikis, get_wiki.

05

Position diffs over time

The structural moat. Show how a founder's view on a topic shifted across two periods, with citations on both sides. Nobody else tracks longitudinal opinion change.

06

Weekly tuning call with me

30 minutes a week to add/remove people, retune topics, surface what your team actually used. Not a CSM — me, the builder.


Don't take my word for it. Click through.

Everything below is live, in production, compiled in the last week. This is the same engine that runs your private instance — just pointed at my personal reading list instead of yours.

Person mega-wiki
Simon Willison
331 compiled entries · v2 wiki
Daily briefing
May 8 PM: Kernels deliver 20-60% LLM gains & Agent brains tackle forgetfulness & RL sims revived
2026-05-08 · 5 min audio + 1500 word article
Topic synthesis
AI Research
Cross-person wiki · pulls from every tracked thinker
Recommendations index
What the smart people endorse
7,680 extracted across the bus
MCP endpoint
JSON-RPC 2.0 for your agents
POST with method: tools/list to see the schema
Full people index
All 2212 thinkers
Every person, sortable by entry count

Your private instance starts here, then we replace this list with yours.

Simon Willison331Garry Tan297Yann LeCun233Jason Calacanis203Andrej Karpathy188Andrew Ng166Ethan Mollick155Tobi Lütke143Jeff Jarvis143Marc Andreessen141

Wide and shallow vs narrow and deep.

A news aggregator tells you what every AI account posted today. That's a feed, not intelligence. absorb.md goes the other direction: a small set of thinkers your fund actually cares about, compiled into a structured graph of claims your team and your agents can query for years.

A

Per-person mega-wikis with cross-referenced claims

Every founder and investor you track gets a living document. Every claim is cited. Every position is traceable to the tweet, podcast, or paper it came from.

B

Claim extraction and contradiction detection

When two thinkers in your bus disagree on RL, synthetic data, or the next architecture, the conflict is surfaced as a structured object, not buried in a paragraph of summary.

C

Multi-source: X, YouTube, podcasts, arxiv, GitHub, blogs

X is one input. We also ingest podcast audio with whisper, YouTube interviews, arxiv papers, GitHub READMEs, and long-form blog posts. Where the actual thinking lives.

D

MCP endpoint your AI agents query directly

Native JSON-RPC 2.0 over MCP. Your internal Claude, Cursor, or custom diligence agent calls the knowledge bus as a tool. No PDF handoff, no scraping, no podcast button to a third party.


Three tiers. Start with a 30-day pilot.
Pilot
$5K
for 30 days
  • 30-day pilot, private MCP endpoint
  • 50 thinkers tracked, tuned to your fund
  • Daily AM briefing (audio + email)
  • Weekly tuning call with the builder
  • Cancel or convert at day 30
or book 15 min first →
Annual
$50K
per year, per fund
  • Unlimited thinkers in your private bus
  • Cross-person disagreement maps
  • Position diffs (longitudinal opinion change)
  • Full agent API over MCP for internal tools
  • Slack integration + AM and PM briefings
  • Weekly tuning call
or book 15 min first →
Enterprise
$100K
per year
  • White-label for LPs and portfolio cos
  • Custom topic slices per practice area
  • Per-partner briefings
  • Portfolio-founder auto-ingestion
  • Dedicated support + monthly strategy review
  • Custom integrations
or book 15 min first →

Compare: one mid-level analyst is $200K all-in. One Bloomberg Terminal seat is $30K/yr and tells you nothing about AI.


Vishal Gurbuxani

20+ years building consumer and B2B platforms. Co-founded Mobclix (Nasdaq IPO via Velti) and Captiv8 ($150M acquisition). Currently at Meta leading AI Evals for Sales AI. I built absorb.md because I was drowning in X trying to keep up with what the smartest people in AI were actually saying. If you're a partner at an AI-focused fund, an AI desk PM, or running corp dev at a F500 with an AI thesis — I want 15 minutes to show you what your portfolio version looks like.

vishal@absorb.md@vgurbuxani on X