If you want to write a weekly AI newsletter faster, stop starting from a blank browser every week. The fastest reliable workflow is to keep your AI sources in one beat-specific Space, get a scheduled digest, ask for the past week's key developments with citations, then turn the shortlist into original commentary.
The goal is not to automate your opinion. It is to cut the collection step so you can spend your time on judgement, selection and writing.
What slows a weekly AI newsletter down
Most of the time goes before drafting:
- checking many sites one by one
- trying to remember what happened on which day
- losing good items in open tabs or saved links
- mixing up signal with noise
- summarising stories before deciding whether they matter to your readers
A manual workflow can work if your source list is small. But if you cover model releases, research papers, developer tools, policy updates and creator commentary, collection becomes the whole job.
Use one weekly workflow
Here is the simple structure:
- Keep one Space for your AI beat.
- Add the sources you actually use.
- Send yourself a weekly email digest from that Space.
- On writing day, ask for the past week's important stories with cited sources.
- Choose stories by reader relevance, not by volume or hype.
- Draft your issue with your own commentary.
This works because every step narrows the pile:
- Follow gathers sources into one place.
- Read and summaries help you triage.
- Ask gives you a cited shortlist for a specific time range.
- Write helps you turn that shortlist into a draft in your own voice.
Set up one Space for your AI beat
A Space is a group of sources for one topic, client or project. For a weekly AI newsletter, make one Space just for the beat you cover.
If your newsletter is broad, one Space may be enough. If you cover a narrower slice such as AI for software teams or open-source models, keep the Space focused so your weekly review stays useful.
Add sources in whichever way is quickest:
- paste a link and let Repurai find the feed
- search by topic
- bulk-import links
- import an OPML file from another reader such as Feedly or Inoreader
Supported sources include:
- Substack
- Medium
- beehiiv
- Ghost
- WordPress and other blogs
- any RSS or Atom feed
- YouTube channels
- Hacker News
- subreddits
- GitHub releases and repositories
- Google Scholar
- arXiv
- Bluesky and Mastodon
Repurai checks followed sources automatically about every hour, and you can refresh manually too.
A good starting source mix for an AI newsletter
Use a balanced mix so your issue is not just product launch notes:
| Source type | What it adds | When it helps most |
|---|---|---|
| Company blogs and Substacks | launches, product updates, framing | catching official announcements |
| GitHub releases and repositories | shipping details, version changes | developer-facing issues |
| arXiv and Google Scholar | new papers and technical claims | research roundups |
| YouTube channels | demos, interviews, explainers | context and reactions |
| Hacker News and subreddits | what people are discussing | spotting reader interest |
| Bluesky and Mastodon | fast expert reactions | early angles and debate |
Cut noise before it reaches your writing session
Fast newsletters depend on good source hygiene.
In your Space, use feed filters to hide items you nearly never use, such as:
- job posts
- podcast episode transcripts if you only want written analysis
- very short link posts
- posts from an author you do not cover
- repeated release notes from low-priority tools
You can filter by keyword, author, link or length, across everything, per Space or per feed, without deleting the source.
This matters because the best weekly workflow is not finding more stories. It is seeing fewer irrelevant ones.
Send yourself a weekly digest
Before your writing day, schedule an email digest for the Space.
Choose:
- the Space for your AI beat
- weekly delivery
- the day and hour you want it
- whether it goes only to you or also to collaborators
The digest opens with an AI-written overview of the stories and links to the source posts. That gives you a first pass before you even open the app.
A good pattern is:
- digest on Friday evening if you draft on Saturday
- digest early Monday if your newsletter goes out on Monday or Tuesday
Use the digest as triage, not as final copy. It should tell you where to look, not what to publish.
Ask for the past week's developments with citations
This is the step that replaces hunting through tabs.
Open Ask, limit it to your AI Space, and set the time range to past week. Then ask for a structured roundup with cited sources.
Use a prompt like this:
Copyable prompt: weekly AI newsletter shortlist
"From my AI newsletter Space, list the 10 most important stories from the past week. Group them into launches, research, developer tools and policy. For each item, give:
- a one-sentence summary
- why it matters to newsletter readers
- one caution or open question
- the cited source post Keep it concise and only use my sources."
Because Ask answers from your own sources and cites the posts, you can move from shortlist to checking the originals quickly.
If the first answer is too broad, ask follow-ups such as:
- "Which three stories would matter most to AI product teams?"
- "What did multiple sources cover this week that still feels underexplained?"
- "Which stories are mostly hype and which include evidence or shipping details?"
- "Which source posts mention numbers, benchmarks, release dates or repository links?"
That gives you not just a list of stories, but a list of stories you can actually write about.
Choose stories by reader relevance
A faster workflow still needs an editorial rule. The usual mistake is choosing the biggest stories on the internet instead of the stories your readers most need help understanding.
Use this test for each candidate story:
1. Is it new to your readers?
If everyone in your niche already saw it, your value is the explanation or the connection, not the announcement itself.
2. Does it change what they do?
Prefer stories that affect:
- tools they use
- workflows they follow
- risks they need to watch
- ideas they should ignore or investigate
3. Can you add a real angle?
Good newsletter commentary often does one of these:
- compares a new release with last week's claims
- translates a research paper into practical impact
- points out what the launch page leaves unclear
- connects separate signals into one trend
4. Can you verify the core claim quickly?
If the original source is thin, the story may be better left out than repeated.
A useful weekly issue usually has:
- 1 lead story with your strongest analysis
- 3 to 5 shorter items
- 1 pattern, lesson or question tying the week together
Draft the issue without losing your voice
Once you have the shortlist, move to Write.
You can create a draft from an article, an idea or a chat answer. For newsletter writers, the easiest path is often to draft from the Ask answer once your shortlist is settled.
Choose Substack if that is your publishing home, or another supported format if you draft elsewhere first. Add Your Voice from a few pasted posts so the draft sounds closer to your normal style, and set post instructions for standing rules such as:
- British English
- no emoji
- short paragraphs
- sceptical tone, not promotional
- explain technical terms once
Then ask for a draft that uses the shortlist as raw material, not as final text.
Copyable prompt: issue draft
"Write a Substack draft for this week's AI newsletter based on these shortlisted stories. Start with the single most consequential item, then cover 4 shorter items, then end with one takeaway about the week. Keep the tone analytical and plain. Do not overstate claims. Use the source material for facts, but make the commentary original and forward-looking."
The point is speed with originality. Let the draft handle structure and transitions, then do the thinking that only you can do:
- cut weak items
- tighten claims
- add your judgement
- add context from previous weeks
- make the intro and ending sound fully yours
If you publish often, the duplicate check is useful before sending the issue. It can warn when a draft is close to something you already published.
Illustrative example issue outline
Here is a simple example you can copy and adapt.
Illustrative weekly AI issue outline
Subject: 5 AI stories worth your time this week
Opening note
- One paragraph: what this week revealed overall
- Example angle: lots of launches, but only two looked likely to change team workflows soon
Lead story
- What happened
- What the source claims
- What seems genuinely new
- What remains unclear
- Why your readers should care now
Story 2: Research
- One paper or result worth noticing
- Plain-language explanation
- Practical implication
Story 3: Developer tool or release
- What shipped
- Who it helps
- Friction, limits or missing details
Story 4: Policy or platform move
- The change
- Who may be affected
- What to watch next
Story 5: One undercovered signal
- Smaller item from your tracked sources
- Why it may matter more than the week's loudest headline
Closing takeaway
- One short section tying the stories together
- Example angle: the useful divide this week was not open versus closed, but shipping detail versus vague promise
That structure is fast because it forces prioritisation.
A practical weekly schedule
If you want the process to feel lighter, assign each task to a fixed moment.
During the week
- save promising posts to read later
- skim summaries to triage quickly
- ignore anything you cannot yet place in a useful category
The day before writing
- read the weekly digest
- open your Space in Magazine, Card, Article or List view, whichever helps you scan fastest
- save 10 to 15 possible items
Writing day
- run the Ask prompt for the past week
- compare the answer with your saved items
- open and verify the original posts for your finalists
- generate a draft in Write
- revise into your own issue
When this workflow works best
This method is best if:
- you cover one beat repeatedly
- you care about being timely
- your issue depends on many public sources
- your value comes from selection and commentary
It is less useful if your newsletter is mostly reported original interviews or mostly long-form essays that do not depend on the week's published posts.
Limits
A fast workflow still has limits.
- AI summaries help triage, but you still need to check original claims before you publish.
- YouTube summaries depend on a transcript, and some videos do not have captions available.
- Ask works from your own tracked public sources, so if a key source is missing from your Space, it will not appear in the answer.
- A weekly digest helps you scan, but it does not replace your editorial judgement.
- Some features depend on plan, so check /pricing if you need a specific workflow.
Do it with Repurai
If you want the shortest path, this is the setup:
- Create one Space for your AI newsletter beat.
- Import your existing feeds or paste links to add sources.
- Add filters to hide recurring noise.
- Schedule a weekly email digest for that Space.
- On writing day, use Ask with the Space limited to past week.
- Turn the cited shortlist into a Substack or newsletter draft in Write.
- Edit for judgement, nuance and originality before publishing.
You can start here: /signup
If you want your sources, summaries and drafts in one place, you can try Repurai free.
FAQ
How many sources should a weekly AI newsletter track?
Track enough sources to cover your beat, but not so many that triage becomes the whole job again. For many writers, a focused list of high-signal sources is better than trying to follow every AI site.
Should I include only the biggest AI stories each week?
No. The best weekly issues usually mix one or two major stories with smaller items that matter more directly to your readers. Relevance beats raw prominence.
Can I use AI to draft the whole newsletter?
You can use AI to speed up collection, triage and first-draft structure. The final issue is better when you add your own judgement, verify the source claims and write the commentary that readers actually subscribe for.