Paul Rohr's Avatar

Paul Rohr

@pevohr

Dad, startup guy. Ideas matter. Design matters. It's about we, not me. hachyderm.io/@pevohr

696
Followers
691
Following
2,756
Posts
28.05.2023
Joined
Posts Following

Latest posts by Paul Rohr @pevohr

Preview
at://sive.rs/app.bsky.feed.post/3mgrtkaazdk2e

FYI. This post links to two different URLs, presumably on purpose?

body, facet = normal page (works)
embed = MP4 (404s)

pds.ls/at://did:plc...

11.03.2026 12:37 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Joyful painting of four Black children at play, cartwheeling and jumping

Joyful painting of four Black children at play, cartwheeling and jumping

On Harriet Tubman Day, one of my all-time favorite images by Jacob Lawrence: the joyful & ebullient screenprint "Play," based on panel #4 from the 1940 "Life of Harriet Tubman" series. The series, thirty-one paintings in all, was one of Lawrence's first major bodies of work

10.03.2026 12:55 πŸ‘ 159 πŸ” 51 πŸ’¬ 3 πŸ“Œ 0

Olympic curling: elaborate sweeping operation to tweak the path of the thrown rock, constant screaming, plus an earth-shaking controversy about tapping the rocks post-release to influence direction
Paralympic curling: just huckin it down there, don't see what the big deal is, works just fine

09.03.2026 13:42 πŸ‘ 441 πŸ” 54 πŸ’¬ 7 πŸ“Œ 2

In 2024, of 161.4 million registered voters, which constituency was responsible for:

- 0.0008% of the votes, +
- 19% of the campaign spending?

A tiny, wildly over-represented group of *300* billionaires

Why should any of the rest of us vote for candidates unwilling to repeal #CitizensUnited?

10.03.2026 16:43 πŸ‘ 2 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

The proper way to translate this, if you want to assimilate the value of America's higher education sector into brute economic terms, is:

"Exports of one of America's world-leading industries dropped by more than a third this year."

Where's that headline in the business press?

10.03.2026 15:34 πŸ‘ 81 πŸ” 34 πŸ’¬ 2 πŸ“Œ 4

He knows he can't say the truth, but it's slipping out anyway and that's the whole thing; "DEI"β€”like wokeness, like critical race theory before it, and political correctness before thatβ€”has always and only ever meant "establishing equality by leveling traditional ascriptive hierarchies"

10.03.2026 14:54 πŸ‘ 24 πŸ” 10 πŸ’¬ 1 πŸ“Œ 0

Gotcha. If you find a way that makes sense, great. If not, that in no way undermines the impact of everything you've already accomplished

Thanks so much for all of this!

10.03.2026 15:38 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

But if you ever want to expand the scope to help us understand WHO they are + what they're like, two useful additions might be:

- a photo on each profile page
- links to video alternatives for text sources like this

Adding faces to names + watching them in action really brings these folks to life

10.03.2026 15:25 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

I love how meticulously your site documents what these DOGE folks have done where

As is, it's an invaluable resource + I'm sure that staying tightly focused on "what, where, + names" has been a huge factor in your success

10.03.2026 15:25 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

I thought I did too, but was surprised by the additional emotional impact of watching him confirm the story you've been meticulously reconstructing all along

you = laying out facts in black + white

him = attempting to calmly cover the same ground as if his nonsense was NBD

Packs a wallop

10.03.2026 14:22 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

Always interesting listening to the guilty tip-toe around depositions.

It's everything we suspected.

They grabbed a bunch of tech-bros referred by Elon and silicon valley investors, set them free in the government, no thought about organizational methods and let them run wild.

10.03.2026 02:25 πŸ‘ 22 πŸ” 8 πŸ’¬ 2 πŸ“Œ 1

I assume the most efficient way for you to update dogetrack.info/wreckers/nat... etc. was to skim the deposition transcripts to confirm or augment details you'd already found from other sources

Have you also had a chance to watch any of the corresponding videos, or was that too much to stomach?

10.03.2026 13:57 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Preview
How Standards Are Made Recently I received a large grant from Bluesky Social PBC to fund my work on FedCM for decentralized web. So whilst the response has been overwhelmingly positive, there's currently a tiny tiny percent...

I wrote up a piece on how standards are made, because of some folks being very negative about the grant on other social networks: writings.thisismissem.social/how-standard...

10.03.2026 03:59 πŸ‘ 158 πŸ” 37 πŸ’¬ 6 πŸ“Œ 3

Don't forget that CEOs report to their boards. Bluesky's board is Jay Graber, Jeremie Miller, Kinjal Shah, and Mike Masnick.

09.03.2026 20:53 πŸ‘ 35 πŸ” 7 πŸ’¬ 1 πŸ“Œ 0

... while also retaining the option to pull a Cheney at the end of the search, or a Jobs return down the road if needed πŸ˜‰

For now, I certainly won't second-guess the strategic wisdom of the woman who did what it took to foster this sandbox we all have the luxury of playing in!

10.03.2026 02:05 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

So long as she retains her role as board chair during + after the search for her Schmidt equivalent, she retains more leverage than most VC-backed founders

And we can start seeing what happens when she's freed to focus on working more directly with the team to advance her vision for the Atmosphere

10.03.2026 02:05 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

I always hate to see startup founders handing over the reins to a "more experienced" CEO -- doubly so when that founder has more to offer than most cis white males, including a proven ability to negotiate the right governance model for their organization

But this is a *very* good sign

10.03.2026 02:05 πŸ‘ 4 πŸ” 0 πŸ’¬ 2 πŸ“Œ 0

I think @jay.bsky.team's biggest accomplishment: she allowed the Bluesky engineering team to follow their hearts on what they saw as pursuing the best design. Not easy to do given all the other pressures in the role.

Best of luck with what's next, and I hope future leadership does the same.

09.03.2026 20:39 πŸ‘ 102 πŸ” 10 πŸ’¬ 1 πŸ“Œ 2
Post image

I've spent the past couple months hand crafting a new B-tree map implementation for Go. It's fast. Faster than Rust's BTreeMap. Faster than the fastest C++ (frozenca/btree). And about 2x faster than my current Go btree (tidwall/btree).

09.03.2026 23:42 πŸ‘ 105 πŸ” 7 πŸ’¬ 5 πŸ“Œ 0

Bravo!

In a world which keeps valorizing the lowest-common-denominator approach of using LLMs to churn out boilerplate code, it's so refreshing to see folks working hard to actually *advance* the state of the art

If we as a profession stop celebrating real accomplishment, we've lost the plot

10.03.2026 00:21 πŸ‘ 2 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

I wasn't expecting to see both phenomena happening so consistently within a few days across every single mushroom

... but I guess with a large enough N, anything can happen!

10.03.2026 00:09 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

Sorry. I was muddling the roles of various tap modes vs. the relay's pure firehose

So then is it safe to assume that:

A. any accountCount increases are organic (first account activity since relay started)

B. any decreases are a result of accounts going inactive (for whatever reason)

10.03.2026 00:01 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

As mentioned in the quoted 🧡, all these stats are per-mushroom averages within time-based cohorts, but when I look at the raw numbers for each individual mushroom, the patterns are strikingly similar

There's far less variability within cohorts than between them

09.03.2026 23:33 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

But why the background increase for much earlier accounts?

Does that always happen? Do those represent accounts with zero activity since the sync1.1 relay started -- say, folks who suddenly resurfaced because of today's announcements?

Or is might this just be an artifact of ongoing backfill?

09.03.2026 23:30 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

With fresh accounts flooding into the newest cohort, significant (seq, accountCount) increases make a ton of sense

Likewise, I wouldn't be surprised if there were enough bans among the next-earlier cohort to *decrease* the accountCount there

09.03.2026 23:30 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

But accountCount growth is harder to explain:

2023-11: 0.05% 95 elders

2024-01: 0.18% 207
2024-09: 0.15% 177
2024-10: 0.15% 268
2024-11a: 0.14% 232
2024-11b: 0.19% 245
2024-12a: 0.19% 245
2024-12b: 0.17% 210
2025-02: 0.17% 245

2025-04: 0.00% 24
2025-08: 0.00% -5 !!
2026-02: 31.78% 38272 newbies

09.03.2026 23:30 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 1

Change rate between seq snapshots is consistent:

2023-11: 0.47% (elders)

2024-01: 0.34%
2024-09: 0.43%
2024-10: 0.43%
2024-11a: 0.42%
2024-11b: 0.41%
2024-12a: 0.39%
2024-12b: 0.43%

With recent cohorts growing fastest:

2025-02: 0.52%
2025-04: 0.81%
2025-08: 1.61%
2026-02: 45.77% (newbies)

09.03.2026 23:30 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

@bnewbold.net Do you know whether the new relay is fully synced with all of the mushrooms, or still backfilling?

Got some unexpected results doing a cohort analysis comparing com.atproto.sync.listHosts snapshots from today vs. earlier this weekend

09.03.2026 23:30 πŸ‘ 1 πŸ” 0 πŸ’¬ 2 πŸ“Œ 0

Some very cool work last week by @baileytownsend.dev on characterizing non-bsky PDSes inspired me to take a longitudinal peek at mushroom PDS usage

If you like geeking out about stuff like this, be sure to scroll upthread to Bailey's take for a more provocative analysis than my modest contribution

09.03.2026 20:59 πŸ‘ 2 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

Emelia has already done so much great standards and auth work behind the scenes, which benefits not just atproto/atmosphere, but also ActivityPub, Solid, and other social web ecosystems.

really excited about the potential for FedCM to make login flows slick and competitive with Google/Apple/Meta

09.03.2026 19:43 πŸ‘ 103 πŸ” 24 πŸ’¬ 2 πŸ“Œ 0