Menu
Sign inAccount
Live

This is yourpulse.

The important thing is not to stop questioning. Curiosity has its own reason for existing.
Albert EinsteinLIFE, 2 May 1955

127newsrooms in the base1,205live stories2,998publications in 24 hours

An event, not a note

Twenty outlets write about the same thing. Pulse shows it as one card and tells you how many sources have written and how fast it spread.

The score is in the open

The order of the feed is a formula, not an editor's taste. It is broken down into multipliers below, and you can turn the dials and watch what changes.

The link goes to the author

We do not reprint anyone's text. A headline, a short summary, a time — and a way through to the outlet that wrote it.

How the feed works

Every number behind the order, and every source it comes from.

How a story gets here

Five steps from someone else's feed to a place in Pulse — every five minutes.

01

Collect

Every five minutes we walk the newsroom feeds and public Telegram channels and take what we have not seen yet.

every 5 minutes
02

Cluster

Publications about one event become one story. The vote is counted per newsroom: an outlet rewriting itself three times still gets one.

1 outlet = 1 vote
03

Tag

A language model reads the headline and the first sentence and assigns topic, geography and how heavy the content is. Once per story — so nothing flickers.

once per story
04

Score

Reach, speed, vocabulary burst, attention on Telegram, reader votes and freshness — in five windows at once, each with its own gravity.

5 windows
05

Feed

A story competes only inside its window. The 30-minute feed rewards what is spreading right now; the month rewards what turned out to matter.

no editor in the loop

The formula, and a dial for it

This is the formula the feed is running on right now. Move the sliders and watch what actually moves the score — and what barely does.

Window
192story score×44 vs. a lone note in the same window
  • Reach×10
  • Lone-source penalty×1.00
  • Speed×1.93
  • Vocabulary burst×1.00
  • Telegram attention×1.00
  • Reader votes×1.00
  • Freshness×0.10

Five windows

The same event lives in five feeds at once, with a different score in each. The difference is how fast news gets old.

0½1 · time, in window lengths100% · score left

What the AI does

Four jobs, and none of them decides what you get to see.

Tag

First light: Hera resolves two asteroids instead of one

SpaceWorldweight 0

Topic

Reads the headline and the first sentence and picks one of four topics: AI, technology, space or science. A story that fits none of them is hidden rather than shown with a borrowed badge: “Elon Musk buys a football club” is not tech, even with Musk in the headline.

How heavy it is

Rates how hard the content is for a passing reader and gives two words for the label. That becomes the sensitive-story veil — a warning, not a verdict.

Story digest

A short retelling on the story page, so you can see what happened without opening eight tabs.

Cross-language merge

Decides whether two headlines are the same event where matching names is not enough: a death and an album announcement share the people, not the event.

What the feed does not do

API

The feed is served by the same response the page itself reads. No key, read only.

Base URL →

Sources under watch

Newsroom feeds

SourceTopicPublications, 30 daysBroken firstWeight
dev.to32,6527,454
Hacker News2,085906
Phys.org2,134784
Engadget775163
The Register485132
TechCrunch532129
Nature413194
Хабр — новости1,407801
iXBT2,4831,354
3DNews1,108470
Phoronix32196
CNews1,8621,558
Product Hunt538412
InfoQ17554
MIT Technology Review10444
OpenAI7317
Nature News7619
Tom's Hardware5212
Science7925
Scientific American4210
Cloudflare Blog6828
NVIDIA466
Electrek5317
Live Science4318
Space.com499
ScienceDaily4816
Stack Overflow Blog336
Ars Technica — ИИ230
The Verge — ИИ192
Simon Willison7545
WIRED — наука210
Latent Space233
MIT News4615
Hugging Face468
JetBrains5312
Google DeepMind115
IEEE Spectrum4021
Universe Today2813
Quanta Magazine153
The Decoder162
Science News2111
SpaceNews162
GitHub Blog204
Netflix Tech Blog166
MarkTechPost160
AWS Machine Learning235
Google Research124
Хабр — Искусственный интеллект7861
ScienceAlert165
WIRED — ИИ110
Хабр — лучшее231194
Figma Blog188
Ars Technica — космос222
Docker Blog110
NASASpaceflight114
Hackaday158
N+1234216
Naked Science220207
Ars Technica — технологии200
Google AI50
NASA186
AWS News141
Системный Блокъ74
Хайтек+4537
Python Insider185
NASA JPL111
Go Blog52
PostgreSQL3321
EarthSky127
Ars Technica — наука220
Martin Fowler108
Nautilus178
Nielsen Norman Group85
Sebastian Raschka20
Import AI41
Stripe Blog94
ESA85
Smashing Magazine51
The Planetary Society131
Node.js119
ПостНаука104
Interconnects61
Astronomy Magazine127
NASA — МКС114
Индикатор2116
Spaceflight Now120
Microsoft Research95
Rust Blog55
Smithsonian — наука112
CSS-Tricks21
Хабр — Космонавтика204
ESO10
Chrome Developers41
Элементы41
A List Apart00
Last Week in AI30
Lil'Log00
Mozilla Hacks00
One Useful Thing20
React00
VentureBeat AI00
web.dev10

Telegram

SourceTopicPublications, 30 daysBroken firstWeight
Machinelearning16680
Rozetked265100
Data Secrets14756
iGuides370150
эйай ньюз3511
Библиотека программиста8661
Типичный программист10774
Denis Sexy IT4619
TechSparks3627
gonzo-обзоры ML статей3929
Стартап дня4035
Сиолошная3417
Веб-стандарты3014
Будни разработчика3528
addmeto21
Техасский Вестник1710
АйТиБорода61
Научная Россия2016
Валера Ковальский197
AI для Всех107
Роскосмос129
ТЕХНО: Яндекс про технологии164
ML Underhood31
Малоизвестное интересное154
Dev Easy Notes00

Suggest a source

A feed or a public Telegram channel. We check what it actually serves and, if it fits, it joins the feed — the ranking treats it the same as the rest.