How It Does What Now? Organizes Its AI Timeline
Updated
It Does What Now? is a chronological, source-linked timeline of artificial intelligence from January 2020 to the present, organized in two layers: a year-by-year spine (2020 through 2026) with dated entries inside each year, and a set of topic "tracks" you can use to filter those entries. Every brief links to its original source, and you can switch between detail levels to control how much you read. It suits anyone who wants to look up what happened in a given year or follow one thread — models, money, policy, accidents — without reading the whole era in order.
The two organizing axes
1. Time: years, then dates
The main structure is strictly chronological. The page shows year sections from 2020 to 2026, and within each year, entries are listed by date, newest first at the top of the current year. The site describes its scope as "January 2020 — September 2026, and counting," with new entries added as news happens.
Each entry is a short brief rather than an article. For example, the 29 Sep 2026 entries include:
- Anthropic finds a Chinese open-weight model can build working cyber exploits
- China says its generative AI users passed 700 million
- OpenAI launches always-on personal agents at its DevDay conference
- Trump orders agencies to call AI "super intelligence" and launches an AI government portal
- AI firms and building unions form an alliance on data-centre standards
- AMD agrees to buy Fei-Fei Li's World Labs for $8.2bn
So a single day can carry several unrelated stories, each tagged and linked separately.
2. Topic: filterable tracks
Alongside the timeline, entries are classified into tracks. The site lists these with entry counts, which tells you both what categories exist and roughly how much material sits in each:
| Track | Entries |
|---|---|
| Models & capabilities | 524 |
| Safety & alignment | 267 |
| Money & business | 268 |
| Government & policy | 232 |
| Open weights & ecosystem | 212 |
| Culture & impact | 196 |
| Security & misuse | 181 |
| Benchmarks & progress | 161 |
| Compute & infrastructure | 154 |
| Ideas & essays | 127 |
| Courts & copyright | 126 |
| Labs & people | 106 |
The counts are a snapshot, not fixed values — they grow as entries are added. A "Filter by track" control lets you narrow the timeline to one or more of these categories, so you can, for instance, read only policy developments across all years, or only safety and alignment items.
Controlling how much detail you see
The page offers several density levels rather than one fixed view:
- Headlines — the shortest form
- Major
- Notable
- Everything — the fullest view
There is also a separate Commentary stream (labeled "what people are saying"), which collects opinion and analysis pieces rather than news briefs. The page shows commentary items with their publication and date, for example:
- Don't Worry About the Vase · 29 Sep — "Astra 6.1 Pulled As Insufficiently Aligned Transformer"
- Forethought · 28 Sep — "A Thousand AI Constitutions: A Short Defense"
- Epoch AI · 28 Sep — "AI is getting cheaper faster than any other transformative technology"
Commentary is a distinct layer from the dated news entries, so you can choose between "what happened" and "what people argued about it."
How to actually use it
If you want a specific year: use the year sections (2020–2026) and read down the dated entries. Each year also has a "year in brief" summary, which is the fastest way to get the shape of a year before drilling into individual days.
If you want a specific theme: apply the track filter first, then read across years. This is the better route for questions like "how has copyright litigation developed" (Courts & copyright) or "what has happened in compute buildout" (Compute & infrastructure), because those stories are spread across the whole period.
If you want to verify something: follow the source link attached to each brief. The site states each entry is "a short neutral brief with links to its original sources," so the timeline is a finding aid, not the primary record.
If you're new: the page suggests scrolling "the story so far — the whole era in one sitting," which means reading the timeline from the beginning at a lower detail level rather than jumping to the latest entries.
What it is not
It is a timeline and index, not an analysis or ranking. Entries are dated briefs with links; the interpretive material lives in the separate Commentary stream. The site also links to a newsletter (hosted on Substack) in its footer, but the timeline itself is browsable without an account — nothing in the page evidence indicates the timeline content is gated.