Sixty words. Then the source.

Research papers, tech news, events and a daily podcast, each distilled to sixty words by a person who read the whole thing. Free on iPhone and Android.

NewsNews card, story timeline open
ResearchResearch card
PodcastPodcast card, chapters open
One card

Everything a card has to earn before it reaches you.

This is a real research card from the app. Not a mockup. The parts are always in the same place, so after a day you stop reading the interface and only read the paper.

Research card, annotated
The source, namedPublisher and authors sit on the card, not behind a menu.
The title, as they wrote itNo rewrite, no adjective added.
Sixty words, never moreWritten to fit by a person, not cut to fit by a machine.
Go deeper, in orderHow they measured it. What the paper doesn't claim. Two related papers.
What can you buildThree directions from this paper, for people who make things.
How you read it

Up for the next. Left for the source. Right to come back.

The app teaches it once, on your first card, and gets out of the way. Every card works the same, in all four verticals.

Swipe up for the next one

Swipe upThe next card.

Swipe left for the source

Swipe leftThe original paper, episode, event or outlet.

Swipe right to go back

Swipe rightBack to the card, where you left it.

What we cover

Four domains. The topics inside them are yours to pick.

In60z covers the fields where the primary source is hard to read and easy to misreport. Everything starts on. Switch off a domain, or go a level down and choose topics.

Space & AstronomyQuantum PhysicsParticle & High-Energy PhysicsChemistry & Molecular ScienceBiology & GeneticsFundamental NeuroscienceEarth & Climate ScienceOceans & Ecology
Clinical trialsGeneticsPublic healthDrugs and therapies
Artificial intelligenceChips and hardwareSystemsSecurityEnergy and materials
FundingLaunchesMarketsStartups
Interests: categories and topics
Research card
Research

The paper, without the parts written for reviewers.

arXiv, OpenAlex and open-access journals, read in full and cut to what was actually shown. Tap a row. This is the GPT-4 technical report as it appears in the app.

How they measured it+
Simulated professional and academic exams, standard language benchmarks, and internal safety evaluations. The report gives the scores. It does not give the model's size, architecture, training compute or data, and the card says so.
What the paper doesn't claim+
That test scores predict reliability in your setting. The authors say it themselves: strong benchmark results do not remove hallucination, inconsistent behaviour, or the need for human oversight. The card keeps that sentence.
What can you build+
Three directions from this paper. An evaluation harness for your own domain before you trust anyone's scores. A review step wherever the report says oversight matters. A retrieval layer that pins every answer to a source you can open.
Related papers+
Two to four papers the editor read alongside this one, each with its own sixty words and its own link.
Podcast card
Podcast

Two hosts. One paper. Nothing the paper doesn't say.

Kavya and Arjun are AI voices, and they work to one rule: Kavya asks what the source doesn't prove, Arjun answers only from the source. This is the episode on Meta's open-weights release, with its chapters open.

Jump to a moment+
  • 01:05
    The claim in one sentenceWhat the organisation actually announced.
  • 04:10
    How the method worksA plain walkthrough without jargon piles.
  • 07:40
    Where it still failsLimits teams hit in practice.
  • 10:20
    What to watch nextSignals that would change the story.

Tapping a moment seeks and plays.

Who they got it from+
  • 01
    Meta AI primary technical noteMeta AI, 2024. Read in sixty words.
  • 02
    Independent explainerScience journalism, 2025.

Every claim in the episode traces to one of these.

The episode in sixty words+
The card you see above the player. If you only have a minute, that is the episode.
Events card
Events

Whether it's worth the trip, before you book it.

Conferences, webinars and hackathons in science and technology, taken from the organisers. This is a student innovation event as it appears in the app.

Who's speaking+
Names and affiliations exactly as the organiser publishes them, refreshed when the agenda changes. No speaker is listed on our say-so.
Getting there and costs+
Venue or link, format (in person, live, recorded), the fee, and the start time already converted to your zone.
Register+
The organiser's own registration page, opened in full. Never a reseller, never a form of ours in between.
This week, by type+
On the home screen: upcoming events split into conferences, webinars and hackathons, with recorded sessions marked so you know what you can catch later.
News card
News

What happened, what isn't confirmed, and where the outlets disagree.

Written from the original reporting and attributed to it. When credible outlets disagree, the card says where. The News tab lists the day's stories by topic; tap a row to open a card on an AWS announcement, sourced to Reuters, AP and Bloomberg.

Where sources differ+
The three outlets agree on the announcement itself. They differ on the timeline and on how much is still unreleased. The card names the disagreement instead of picking a side.
What isn't confirmed+
Full evaluation numbers. Detailed benchmarks were not in the initial note and remain unpublished. Until they are, the card treats the performance claim as a claim.
Story timeline+
  • 08:10
    Press note postedThe company publishes the announcement.
  • 09:40
    Wire stories landReuters and AP circulate first writes.
  • 11:05
    Analyst notesMarket coverage debates the impact.

All times in your zone.

Follow this story+
One tap to keep a story on your list. When it moves, you hear about it once, not every time someone rewrites it.
Listen

A short episode from the day's research.

Kavya and Arjun are AI voices. Each episode takes one paper and works through it in chapters: the claim, the method, where it fails, what to watch. Every claim traces to a named source.

Meta AI · Kavya & ArjunIn the app

Open weights and the LLaMA release model

What releasing model weights actually opens up, and what it doesn't: access, customisation and practical trade-offs, and the questions openness raises about responsibility, evaluation and deployment.

Episode audio is in the app today. It arrives on the site with launch.
  1. 01:05
    The claim in one sentenceWhat the organisation actually announced.
  2. 04:10
    How the method worksA plain walkthrough without jargon piles.
  3. 07:40
    Where it still failsLimits teams hit in practice.
  4. 10:20
    What to watch nextSignals that would change the story.
SourcesMeta AI primary technical note · Independent explainer

A set, not a feed. It has a top and a bottom.

Discover shows what's across In60z today, by domain, then the week's events. You choose how many cards arrive each morning and at what time, and which categories and topics they come from.

Discover: today across In60z

DiscoverToday across In60z, by domain.

Discover: upcoming events

Further downThe week's events, recorded ones marked.

Research tab

Each verticalIts own list. "All 4 research today", then it ends.

Three promises, kept on every card.

60
Never more than sixty words

A hard limit, enforced where the card is stored. Not a style guide. If a claim needs more, it gets a source link, not a second paragraph.

1
A person reads every card before you do

Machines draft. Editors publish. Anything the source doesn't support is cut, and any language that sells rather than tells is rejected. If nothing clears the bar, nothing runs.

↗
Every card links its source

Original writing, attributed, with the publisher's own page one tap away. Nothing is republished and nothing is taken from behind a paywall.

The source, opened from a card

Read today. Then get on with it.

iPhone and Android.