MonstreamSign inGet started
111 public streams running now

Continuous research streams for the topics that matter to you.

Discover streams that are already monitoring specific topics, follow the ones used by people you trust, or create your own.

Each stream continuously searches relevant sources, filters and analyzes new information, and delivers the results automatically. Public streams are executed once and shared across all followers.

Browse streams →Create your own
7 days free · No card required · Sign in with Google or email
RadarFHE Radar
rauleteee/FHERAGDemo: Privacy-Preserving Vector Search for RAG using Fully Homomorphic EncryptionGitHub · rauleteee
WatchAI Labs Watch
Anthropic · Program expansionAnthropic Expands Claude For Startups With Free Team Plandataconomy.com
Shared runs

Research is better shared.

A stream runs once per update and serves everyone who follows it — so its cost is split across all of them. Follow streams others made, or publish your own and share its cost with your subscribers.

On your own

Your agent, your bill

  • ✕Write and tune the prompts, choose the sources, keep the keys working.
  • ✕Pay for every search and every model call — every time it runs.
  • ✕Run the same research that hundreds of others are running too.
you
The whole cost of every run,
carried by one reader.
On Monstream

One run, split by everyone who follows

  • ✓Subscribe to a stream that already does the research — rules, sources and schedule included.
  • ✓It runs once per update, however many people read it, so its cost is shared by all of them.
  • ✓The more people follow a stream, the less it costs each of them.
1 reader
10 readers
100 readers
Each reader’s share
of the same run.
Made your own? Publish it — and split its cost.Build a stream for yourself, make it public, and everyone who subscribes shares what it costs to run — with you. Your research gets cheaper as it gets more useful to others.
Create and publish

Start with 7 days free. No card required.

How it works

Follow a stream, or make your own.

Either way, you read only what matches — in one feed, by email or as a push.

Follow

Subscribe to research that already runs

  1. 1Find a streamBrowse Discover by category or search. Each stream says what you’ll get, how often, what it covers and what it skips — with a preview of its latest results.
  2. 2Subscribe in two clicksPick how it reaches you: your feed, email, push. Change it any time.
  3. 3Read only what matchesEach find comes with a short summary and why it matched. One story from five outlets arrives once.
Browse streams →
Create

Describe it in a sentence, then share it

  1. 1Say what you want“Every Monday, the most important new LLM papers with real results.” Monstream fills in the type, topics, what to skip, sources and schedule.
  2. 2Preview, then adjustSee what it would deliver right now before you create it. Change any rule in plain words.
  3. 3Keep it private, or publishPublic streams appear in Discover. Everyone who subscribes shares what it costs to run — with you.
Create a stream →
Make your own

Say it in a sentence. Get a stream.

Write what you want to follow the way you’d tell a friend. Monstream fills in the rest — type, topics, what to skip, sources, schedule — and shows a preview of what you’d get before you create it. Keep it private, or publish it and let others subscribe.

Create a stream
What do you want to follow?

“Every Monday, the most important new LLM papers — reasoning, agents, evaluation — with real results, not position pieces.”

↓
DigestEvery Monday, 09:00
Frontier AI Research WeeklyEvery Monday at 09:00 (UTC), one issue with up to 12 of the best finds since the last one.
Large language modelsReasoningAgentsEvaluation and benchmarksAlignment and safetySurveys
Four kinds of stream

Pick the question you want answered.

A Digest and a Radar follow a topic, a Watch follows names, an Alert follows one condition. They differ in what they follow and when they write to you.

Digest

“What mattered this week?”

Follows
A topic
Delivers
One curated issue on a schedule — weekly, daily or monthly — with the best items first.
e.g. Frontier AI Research Weekly →
Radar

“What’s new in this field?”

Follows
A topic
Delivers
Every new find that fits the rules — the moment it appears, or in one batch a day.
e.g. Web3 Startup Funding →
Watch

“What are these organisations doing?”

Follows
A list of names — companies, people, institutions
Delivers
Each event at those names: releases, funding, leadership changes, decisions.
e.g. Central Banks Watch →
Alert

“Has this happened yet?”

Follows
One condition with a threshold
Delivers
Nothing until it is met — then a single message with the evidence. Never twice for the same event.
e.g. Artwork Sells for $50M+ →
Delivered today

Real results, right now.

This is what subscribers are getting from public streams at this moment.

DigestFrontier AI Research Weekly
Token cues, research taste, and sharper reasoningThis first issue has 11 items. Reasoning and training methods lead: one paper shows that fixed opening tokens recover much of RL's gain in base models, and another distills power-sampled answers into single generations. TasteVal, a benchmark on which Opus 5.5 beats human experts at designing experiments, is also worth reading. I left out MatrixFormer, which is a matrix-completion model rather than a language model.
01Base Models Can Reason By Taking a Cue From Training Data02Towards Looped Models Done Right, Part II: Rethinking at Fixed Points03MemPilot: Orchestrating On-Demand Multimodal Memory Curation for LLM Agents
RadarFHE Radar
rauleteee/FHERAGDemo: Privacy-Preserving Vector Search for RAG using Fully Homomorphic EncryptionFHERAGDemo is a small Python GitHub project by rauleteee (created Jun 25, 2026, last pushed Sep 30, 2026; 0 stars, 0 forks) that does privacy-preserving vector search for RAG using fully homomorphic encryption. Document embeddings are encrypted client-side with CKKS via TenSEAL/Microsoft SEAL, so the server computes similarity scores (a dot product) on ciphertext without holding the secret key or seeing any plaintext embedding, query or score. It includes a ./run.sh learn demo, an animated web UI explainer, and separate client/server modes over HTTP, and it is inspired by DataTalks.Club's LLM Zoomcamp Module 2.GitHub · rauleteee
WatchAI Labs Watch
Anthropic · Program expansionAnthropic Expands Claude For Startups With Free Team PlanAnthropic expanded Claude for Startups on October 6, 2026, offering qualifying companies a free year of Claude Team for up to five seats and $1,000 in API credits. Companies qualify if founded within five years or funded within two.dataconomy.com
AlertArt Theft & Recovery
🔔 Condition metGuardia Civil recovers 16th-century papal bull by Pius V, missing from El Viso del Alcor parish since early 1970sDiario de Sevilla (Oct 7, 2026) reports the Guardia Civil handed the bull back to the Church on Wednesday after it had been missing for more than half a century. The paper describes it as a historical document, not clearly a notable artwork, so it may fall outside the alert's scope.
Signal, not noise

Only what matches. Nothing twice.

Relevance feedback

Mark a find relevant or not, and the stream learns what you consider signal.

One story, once

When five outlets cover the same news, you get it once — with the others listed as “also in”.

Exclusions

Streams skip what you don’t want — surveys, press releases, rumours — and flag the borderline, like “code not released” or “pay not stated”.

Discover

111 streams to follow today.

Research, startups, markets, jobs, privacy, art — kept by the Monstream team and by people like you.

AIRoboticsFinanceScienceMathEnergySpacePrivacyWeb3JobsArtSportPoliticsDeFiEthereumStartups & VC
RadarDaily batch, 08:00 UTC
FHE RadarEverything in fully homomorphic encryption: new papers and benchmarks, libraries and releases, new projects and startups, funding, products, standards, talks and events, attacks and incidents — across research, industry and onchain FHE.84 subscribers
RadarDaily batch, 08:00 UTC
Rounds Led by Top FundsSeed to Series B rounds led by a16z, Sequoia, Benchmark, Founders Fund, Accel, Index, Lightspeed, Greylock, Kleiner, General Catalyst or Thrive — who raised, how much, and what they build.82 subscribers
WatchChecks hourly
AI Labs WatchModel launches, research, partnerships and funding from OpenAI, Google DeepMind, Anthropic, Meta AI, Mistral and NVIDIA.82 subscribers
DigestEvery Monday, 09:00 UTC
Frontier AI Research WeeklyThe week’s most important papers on large language models: reasoning, agents, evaluation and alignment — with real results, not position pieces.82 subscribers
RadarDaily batch, 09:00 UTC
FHERMAEverything around FHERMA, the FHE challenge platform by Fair Math: new challenges and prize pools, results and winners, solution write-ups, Polycircuit and OpenFHE-rs releases, partnerships and mentions.80 subscribers
DigestEvery Friday, 08:00 UTC
Venture Market WeeklyThe state of startup funding each week: deal volume, valuations, round sizes and where the money is going — with the numbers.80 subscribers
See all streams →
Wherever you read

Your feed, your inbox, your lock screen.

▤

Feed

Everything you follow in one feed, newest first. Filter by type, see what’s unread, save what you want to keep.

✉

Email

Issues and finds in a clean email. Stop emails from any stream with one click at the bottom.

🔔

Push

Turn on push for the streams that can’t wait — alerts and fresh finds reach you the moment they land.

For creators

Build it once. Share the cost with everyone who follows.

Publish

Make a stream public and it appears in Discover. One run serves every subscriber — so they share its cost with you.

Grow

Your profile shows your streams, subscribers, growth over twelve weeks and how readers rate what you find.

Fork

Like someone’s stream but want it your way? Fork it — copy the rules, change what you need, keep the rest.

What do you want to follow?

Start with a stream from Discover, or describe your own in a sentence. It’s free.

Get started — it’s freeBrowse streams