The Best Bluesky Starter Packs in 2026 (Ranked by What's Still Alive)
The best Bluesky starter pack for you is whichever one in your subject still has people posting in it — and that's the one thing no starter pack directory will tell you. So I measured it. I pulled 12 well-known packs across news, science, tech, art, politics, and publishing, walked all 1,073 accounts inside them, and checked when each one last posted. 68.6% had posted in the previous 30 days. Per pack, that ranged from 97% down to 32%.
Which is the whole problem with "best starter packs" lists. Researchers studying the format counted 335,420 starter packs in its first six months alone, three-quarters of everything I sampled was built in 2024, and every directory indexing them sorts by recency or by nothing at all. So the ranked list below is sorted by the only thing that matters twenty months later: whether following a pack fills your feed or fills it with ghosts.
What Are the Best Bluesky Starter Packs Right Now?
The best pack is the one whose members are still there — so here are the eleven that scored highest. Measured on July 31, 2026 through Bluesky's public API: I fetched every member of each pack with app.bsky.graph.getList, then asked app.bsky.feed.getAuthorFeed for each member's single most recent post. "Still posting" is the share whose latest post is within 30 days. "Joins" is Bluesky's own counter for people who signed up to Bluesky through that pack.
| Starter pack | Topic | Accounts | Still posting | Joins |
|---|---|---|---|---|
| Major news outlets - VERIFIED | News | 62 | 97% | 171 |
| Andrew Dessler's climate science & policy | Climate science | 50 | 88% | 53 |
| UK Politics | Politics (UK) | 149 | 87% | 327 |
| Davis's Art Start | Art | 150 | 83% | 24 |
| INN Network Newsrooms (Part 1!) | Local journalism | 150 | 78% | 42 |
| Publishing News & Views | Books & publishing | 37 | 65% | 106 |
| Future of journalism worldwide | Journalism | 149 | 60% | 25 |
| Civic Tech Practitioners | Civic tech | 60 | 55% | 23 |
| Python authors, speakers, and teachers | Python | 65 | 54% | 32 |
| Data Science, AI and ML System and Infrastructure | Data science | 23 | 39% | 24 |
| Marine Science & Technology (MASTS) | Marine science | 90 | 32% | 24 |
Three things fall out of that table that I didn't expect.
Popularity and liveness are unrelated. The most-joined pack in my entire sample isn't in the table above — DGGverse - Destiny & Politics has 1,940 joins, more than three times the runner-up, and only 47% of its 88 members have posted in the last month. Meanwhile the 97% pack has 171. Joins measure how good a pack was at recruiting people to Bluesky in 2024. They say nothing about whether the accounts inside it are awake in 2026.
Institutional packs age better than personal ones. The top of the table is newsrooms, journalists, and working scientists — people who post as part of their job. The bottom is conference-and-community packs, where a chunk of members joined during the exodus, posted for a month, and drifted back to wherever they came from. Marine science lost two-thirds.
Nothing here is truly dead. Even the worst pack has 29 accounts still actively posting about marine science, which is 29 more than you'd have found on your own. A 32% pack isn't useless — it just means you should pick from it rather than take all of it.
How Do I Find Starter Packs on Bluesky?
You can search them in the app, which most guides on this topic will tell you is impossible. That advice was true once and has been quietly wrong for a while; the packs are right there under a tab. Three routes, in the order I'd try them:
- In-app search. Type your topic and the results screen has a Starter packs tab alongside Top, Latest, People, and Feeds. (In the app's source it sits behind a feature flag and only shows when you haven't applied post filters, so if it's missing for you, that's why.) It matches against pack names and descriptions, which is why searching
photographybeats searchingphotography starter pack— the second phrase isn't in most pack names. - Straight off a profile. Every account has a Starter Packs tab next to Media and Likes. If you already trust someone's taste, this is a better filter than any search ranking.
- Third-party directories. blueskystarterpack.com lists over 76,000 packs, far more than search will show you, ordered roughly by recency. Bluesky Directory and TheBlue.social run similar catalogues. None of them publish anything about who's inside.
One warning about route 1, because it explains a lot of frustration: starter pack search is ranked by text relevance, not by popularity, quality, or freshness. Searching tech put Paul Frazee's "Tech Folks" pack (2 joins) first and a university pack called "Texas Tech-ians" (0 joins) directly beneath it, while the most-joined pack in the entire result set — Civic Tech Practitioners, at 23 — sat down at rank 44. The endpoint behind the tab takes no auth, so paste https://public.api.bsky.app/xrpc/app.bsky.graph.searchStarterPacks?q=tech into a browser and you'll get the same ordering I did.
Why Are Most Bluesky Starter Packs Dead?
Because the format was built for a sign-up wave that ended. To get a sense of the field rather than a hand-picked dozen, I swept ten topic searches — tech, science, art, music, news, politics, history, humor, business, authors — and collected every unique pack returned. That's 2,847 packs, and the distribution is brutal:
| Starter packs with… | Share of 2,847 |
|---|---|
| Zero all-time joins | 79.3% |
| Fewer than 10 joins | 97.7% |
| 25+ joins | 0.6% |
| 100+ joins | 0.3% |
| Created in 2024 | 75.0% |
Median joins: 0. Mean: 2.1. The maximum across the entire sweep was that one Destiny pack at 1,940.
Before anyone quotes 79.3% back at me, two things it doesn't mean. First, joinedAllTimeCount counts people who created a Bluesky account through the pack link — the app renders it as "# people have joined Bluesky via this starter pack!" — so it isn't counting existing users who tapped Follow All. A pack with zero joins may well have been useful to plenty of people. Second, this is what search returns for ten common words, not a random draw from all ~335,000 packs, and the shares move with how deep you page: I re-ran the sweep an hour later and got 2,911 packs at 79.6% zeroes, but a shallower single-page sweep skews toward the popular end and puts the zero share nearer 60%. The ordering of magnitude is solid. The second decimal place is not.
There's a detail in the app's source that makes the sparsity concrete. The line announcing how many people joined via a pack only renders when the count is 25 or higher (joinedAllTimeCount >= 25 in StarterPackScreen.tsx, where the field's own code comment reads "Number of users (always at least 25) who have joined Bluesky using a specific starter pack"). By my sweep, that's 18 packs out of 2,847 — 0.6%. The social proof exists; it's switched off for 99.4% of the network.
I'll also flag a number I decided not to use. Almost every pack I touched, including the 1,940-join one, reports joinedWeekCount: 0 — across the whole sweep exactly one pack had a nonzero weekly count. That could mean joins have essentially stopped, or that the weekly counter is no longer computed. I can't tell those apart from outside, so I'm leaving it out of the argument — but if you see someone quoting weekly starter pack joins, ask them which one it is.
How Do You Tell a Live Pack From a Dead One?
Three checks, and it's the third one nobody does. Before you tap Follow All on 150 strangers, spend two minutes:
- Open the pack and scroll the member list. The app shows recent posts per account. If you're hitting "3mo" and "1y" timestamps in the first ten, the rest won't be better.
- Check the pack's created date against its subject. A November 2024 pack about a 2024 news cycle is a time capsule. A pack that's been updated recently is a different animal — curators who prune are rare and worth following in their own right.
- Don't trust the member count. This one surprised me: the count Bluesky stores for a pack's list doesn't match what you get back when you actually fetch the members. A music industry pack advertises 251 and returns 248; Davis's Art Start says 151 and returns 150. The missing ones are deactivated, deleted, or taken-down accounts still sitting in the stored total. It's the same lazy-aggregate behaviour that makes your follower count wrong, turning up somewhere new.
That music pack points at something else, incidentally. The official cap is "up to 150 people and up to 3 custom feeds" per Bluesky's launch post, and the app enforces it — but the list underneath a pack is an ordinary list, and 42 of the 485 packs I hydrated in one sweep hold more than 150, the largest at 251. The 150 is a client rule, not a protocol one.
If you want the mechanics — creating a pack, the cap, getting yourself added to other people's — that's the complete guide to starter packs. If you're choosing between a pack and a plain list for your own curation, starter pack vs list has the eight-axis version. And if you're after individual accounts rather than bundles, the best Bluesky accounts to follow is the companion piece.
Where Does Agent Sky Fit?
Everything above is one job done by hand: given a pile of accounts in my niche, which are worth following? I built Agent Sky to run that question on a loop.
A pack is genuinely good raw material — 50 to 150 people chosen by someone who knows the subject, which is better targeting than I can buy anywhere. But it's a photograph of a community taken on one day in 2024, and the table above is what a photograph looks like after twenty months. Following all 90 members of that marine science pack means picking up 61 accounts that haven't posted in a month, 30 of which haven't posted in a year or ever. Your feed doesn't fill up, and none of them follow you back, because nobody's home to see it.
I should be precise about which half of that my own tool solves, because it's less than this post might imply. Agent Sky hard-skips candidates with no bio, no display name, no avatar, or zero posts ever, and it works through the rest at a default of a few follows an hour rather than in bursts that get you flagged as spam. What it does not do is filter on recency — the code comments say so outright, because backtesting found that the accounts that look low-quality on those signals convert to follow-backs best. The liveness triage in this post is still a human job. What the tool is genuinely strict about is the other end: it re-reads the live follow graph before undoing anything, so it never drops someone on stale information. If you'd rather prune by hand, unfollowing inactive accounts is the manual version. Free to start, $9/month after that.
The cheapest version of everything here takes ten minutes and no software, though: find the two or three packs in your actual subject, ignore the join count, and follow twenty people from them by hand. About half the packs I checked beat the 68.6% average, so expect to strike out as often as not — which is still a better hit rate than the search ranking will give you.
Quick FAQ
What are the best Bluesky starter packs? The best ones are the packs whose members are still posting — which is not the same as the packs with the most joins. Across 12 well-known packs I checked on July 31, 2026 (1,073 accounts total), 68.6% of members had posted in the previous 30 days, but individual packs ranged from 32% to 97%. The strongest of the set were Major news outlets - VERIFIED by @newseye.bsky.social (62 accounts, 97% active), Andrew Dessler's climate science pack (50 accounts, 88%), and UK Politics by @adambienkov.bsky.social (149 accounts, 87%).
How do I find starter packs on Bluesky? Search the topic in the Bluesky app and open the Starter packs tab in the results — despite what most guides say, packs are searchable in-app now. It matches pack names and descriptions, so searching the plain topic word ("science", not "science starter pack") returns more. Anyone's packs also appear on a Starter Packs tab on their profile, which is the better route if you already trust the curator. Third-party directories index far more packs than search surfaces, but none of them tell you whether a pack's members are still active.
How can I see what starter packs I'm in? There's still no built-in way. The feature request — social-app issue #5493, "Allow users to see what starter packs they've been added to, and option to remove themselves" — has been open since September 26, 2024, and with 43 reactions it sits in the top 1% of that repo's issues by support. Third-party checkers exist that scan for your DID in public pack lists, since all of this data is public.
Do Bluesky starter packs still work in 2026? As a way to find accounts, yes — a good pack is 50 to 150 people hand-picked by someone who knows the subject. As a growth engine for the curator, much less than they did. Of about 2,850 packs I sampled across ten topic searches in July 2026, 79.3% had never brought a single new user to Bluesky, and 75% were created back in 2024. The format peaked with the sign-up waves it was built for.