A daily front page that says why each story ranked.

Tech & AI, World, and Finance, one dated edition a day. Every story links the original source.

Read today's edition See how stories are chosen

Window: 24 hoursThe span of reporting each edition covers.
Categories: ThreeTech & AI, World, and Finance.
Limit: Up to sixPer category and UTC date.
Sources: One click awayOpen the original reporting behind each summary.

See a day on Rough Day

tech & ai

A stricter test for long-running AI agentswire.example

Browser isolation weakness crosses an expected boundaryledger.example

A smaller compiler representation improves tooling workflowsdispatch.example

world

Regional leaders agree to a monitored ceasefire frameworkwire.example

Powerful earthquake disrupts infrastructure across several provincesledger.example

Governments adopt joint plan for drought-strained river basindispatch.example

finance

Central bank holds rates as inflation risks persistwire.example

Major lender raises provisions as credit stress growsledger.example

Payments companies combine in a cross-market infrastructure dealdispatch.example

tech & ai

A stricter test for long-running AI agentswire.example

Researchers released an evaluation that tests whether language models keep their answers grounded across long technical tasks. It covers agents that write code, analyze documents, and operate tools over extended sessions.

AI and machine learningaideveloper tools

Why it ranked: It changes how teams judge whether an AI agent can be trusted with long tasks.

Read at wire.example

Browser isolation weakness crosses an expected boundaryledger.example

A smaller compiler representation improves tooling workflowsdispatch.example

New chip interconnect targets accelerator memory bottleneckscourier.example

Browser vendors align on resilient offline application APIsbulletin.example

Reproducible checks catch scientific data errors earliergazette.example

$curl -H 'Accept: text/markdown' https://rough.day/?category=tech

# Rough Day

Showing tech & ai only.

## 2026-08-04

### tech & ai

1. [A stricter test for long-running AI agents](https://wire.example/…) · wire.example

2. [Browser isolation weakness crosses an expected boundary](https://ledger.example/…) · ledger.example

3. [A smaller compiler representation improves tooling workflows](https://dispatch.example/…) · dispatch.example

Step 1 of 4. One ranked page a day, each story beside its source.

Read the short introduction

How stories are chosen

Collectprevious 24 hours
Gather headlines, descriptions, and excerpts, and group reports about the same event.
Keep the sourcesource link kept
Each story keeps its source link and the attribution and time recorded when it was collected.
Rankconsequence · confirmation · timeliness
Rank by the criteria for each category below.
Publishup to 8 per category
File up to eight stories per category under the UTC date, and fewer when the day does not support eight.

tech & ai

Consequential developments in software, artificial intelligence, security, infrastructure, hardware, science, and technology policy. Selection favors demonstrated impact and durable change over product promotion or speculation.

world

Geopolitics, conflict, diplomacy, elections, humanitarian crises, natural disasters, climate, migration systems, energy-supply shocks, and other events with major regional or cross-border consequences. Selection favors verified scale, urgency, and lasting public impact over isolated local incidents.

finance

Material changes in economies, central banks, markets, banking, payments, regulation, major companies, and globally consequential commodity-supply shocks. Selection favors broad financial consequences over routine price moves, forecasts, or personal-finance advice.

Daily archive

Read the seven most recent dates with editions, newest first, with category filters and source links. The Older dates link at the end of each page goes back seven more.

https://rough.day/

Category link

A link you can share to Tech & AI, World, or Finance alone.

?category=tech

Markdown

Send this header to get the archive, this page, or a news article as Markdown at the same address.

Accept: text/markdown

Atom feed

Add this address to a feed reader to follow new editions. The feed holds the last 14 editions, newest first, each with headlines, summaries, tags, and source links. /feed/finance.xml is Finance only.

https://rough.day/feed.xml

Edition JSON

The site guide lists machine-readable editions for the latest date, a named UTC date, or an index of available dates. Each story includes its summary, why it ranked, tags, tickers when known, article time, and edition time.

https://rough.day/llms.txt

Site guide and sitemap

A guide for AI tools that explains when to use Rough Day, and a list of the public pages.

/llms.txt · /sitemap.xml

What each story shows

Before you open a story: Date, category, rank, title, and source site.
After you open it: Summary, topics, why it ranked, and links to the source.
Limits: Rough Day publishes short summaries in one edition a day. It does not carry full articles or live updates, and it leaves out many important stories.

How Rough Day compares

A daily shortlist, live coverage, or community discussion: choose the pace you want.

Three ways to catch up
	Rough Day	Techmeme	Hacker News
Coverage	Tech, world, finance	Tech	Mostly tech
Format	One short daily edition	Coverage grouped all day	Links and discussion
Selection	Ranked with reasons	Editors and software	Reader submissions and votes

Techmeme and Hacker News describe their formats on their sites.

AINews by smol.ai: A longer weekday recap of AI news and community discussion.
TLDR: Free daily tech summaries by email, with sponsored slots.
Google News: Personalized headlines from many outlets, updated all day.

Sources and privacy

Reading needs no account. Newsletter signup is closed.

Sources and summaries

Each story's link and attribution come from its source. Rough Day chooses each story's rank and topics and writes its title, summary, and the note on why it ranked.

Reader access

Reading needs no account, and the archive works without JavaScript.

Analytics

The site may send anonymous page-view, performance, and error events without cookies or URL query strings. Where consent is required, events wait for your choice; analytics stays off if the region is unknown or you decline. It does not record headlines, source links, who you are, session replays, or general clicks.

Newsletter

Newsletter signup is closed. You can use confirmation and unsubscribe links already sent by email to manage your Rough Day subscription. Reading the archive needs no account.

Common questions

Why can a category have fewer than eight stories?

Eight is a ceiling. When the day's reporting doesn't support eight distinct stories, a category has fewer.

Does archive reading require JavaScript?

No. The server sends the headings, stories, summaries, and links as HTML, and stories open without JavaScript.

How current is an edition?

Each edition covers the previous 24 hours and is filed under its UTC date. Rough Day is a daily archive, not a live feed.

How do I find an earlier day?

Open https://rough.day/archive/YYYY-MM-DD with the UTC date you want. The page shows that date and the six earlier dates with editions, and you can bookmark its address. The archive keeps the 60 most recent dates with editions; a date without an edition, or one that has left the archive, shows a not-found page.

Can another tool read the same page?

Yes. Request a public page with the header Accept: text/markdown to get Markdown. The /llms.txt guide lists the formats and their limits.

How can I follow new editions?

Add https://rough.day/feed.xml to your feed reader for ranked headlines and original source links from the last 14 edition dates.

What does a rank mean?

A rank is Rough Day's order of importance within one category on one date. It doesn't compare stories across days or categories.

Open the current archive.

Start with all categories, or use the shareable Tech & AI, World, and Finance filters.

See the archive Open the site guide