fig (aka:[phil])
@bad-example.com
art and transistors, they/them plant-mom building community infra
🌌 constellation.microcosm.blue
🚒 relay.fire.hose.cam jetstream.fire.hose.cam
🛸 UFOs.microcosm.blue
🎇 spacedust.microcosm.blue
💥 notifications.microcosm.blue
🛰️ slingshot.microcosm.blue
🌌 constellation.microcosm.blue
🚒 relay.fire.hose.cam jetstream.fire.hose.cam
🛸 UFOs.microcosm.blue
🎇 spacedust.microcosm.blue
💥 notifications.microcosm.blue
🛰️ slingshot.microcosm.blue
Pinned
fig (aka:[phil])
@bad-example.com
· Apr 21
…but for real, you can help keep this alive if you want:
github.com/sponsors/uni...
ko-fi.com/bad_example
github.com/sponsors/uni...
ko-fi.com/bad_example
let repo-stream loose on 173K real atproto repos and it broke 3M records/sec, averaging around 2M/s.
(this is: reading CARs, validating blocks, walking and validating the MSTs, emitting and counting all repo keys)
(this is: reading CARs, validating blocks, walking and validating the MSTs, emitting and counting all repo keys)
November 12, 2025 at 6:30 PM
let repo-stream loose on 173K real atproto repos and it broke 3M records/sec, averaging around 2M/s.
(this is: reading CARs, validating blocks, walking and validating the MSTs, emitting and counting all repo keys)
(this is: reading CARs, validating blocks, walking and validating the MSTs, emitting and counting all repo keys)
Reposted by fig (aka:[phil])
good morning ☀️ happy wednesday. i meant to post yesterday for tuba tuesday but got distracted with work
www.youtube.com/watch?v=-WW4...
www.youtube.com/watch?v=-WW4...
Anthony Braxton: Composition No. 19, SoundSpace at the Blanton Museum
YouTube video by Steve Parker
www.youtube.com
November 12, 2025 at 1:40 PM
good morning ☀️ happy wednesday. i meant to post yesterday for tuba tuesday but got distracted with work
www.youtube.com/watch?v=-WW4...
www.youtube.com/watch?v=-WW4...
some atproto repo size numbers from an ~80k sample, in case you need to plan for idk full network backfill or something
85% of repos are <1MiB (as CARs)
<5MiB: 95%
10: 98%
20: 99.2%
32: 99.7
85% of repos are <1MiB (as CARs)
<5MiB: 95%
10: 98%
20: 99.2%
32: 99.7
November 11, 2025 at 5:18 PM
some atproto repo size numbers from an ~80k sample, in case you need to plan for idk full network backfill or something
85% of repos are <1MiB (as CARs)
<5MiB: 95%
10: 98%
20: 99.2%
32: 99.7
85% of repos are <1MiB (as CARs)
<5MiB: 95%
10: 98%
20: 99.2%
32: 99.7
repo-stream v0.2 is outttttt
- interleaved block loading/walking is out (no point)
- spilling to disk for huge CARs is in!
you can process really large CARs without proportionally growing memory usage, at around 1M records/sec.
tangled.org/@microcosm.b...
docs.rs/repo-stream/...
- interleaved block loading/walking is out (no point)
- spilling to disk for huge CARs is in!
you can process really large CARs without proportionally growing memory usage, at around 1M records/sec.
tangled.org/@microcosm.b...
docs.rs/repo-stream/...
@microcosm.blue/repo-stream
Fast and robust atproto CAR file processing in rust
tangled.org
November 10, 2025 at 8:37 PM
repo-stream v0.2 is outttttt
- interleaved block loading/walking is out (no point)
- spilling to disk for huge CARs is in!
you can process really large CARs without proportionally growing memory usage, at around 1M records/sec.
tangled.org/@microcosm.b...
docs.rs/repo-stream/...
- interleaved block loading/walking is out (no point)
- spilling to disk for huge CARs is in!
you can process really large CARs without proportionally growing memory usage, at around 1M records/sec.
tangled.org/@microcosm.b...
docs.rs/repo-stream/...
Reposted by fig (aka:[phil])
Reposted by fig (aka:[phil])
great thread of our IETF BoF session today
huge thanks to everyone that participated in person & remote
for more info on what a "BoF" is & what are plans at the IETF are check out: docs.bsky.app/blog/taking-...
huge thanks to everyone that participated in person & remote
for more info on what a "BoF" is & what are plans at the IETF are check out: docs.bsky.app/blog/taking-...
November 5, 2025 at 9:10 PM
great thread of our IETF BoF session today
huge thanks to everyone that participated in person & remote
for more info on what a "BoF" is & what are plans at the IETF are check out: docs.bsky.app/blog/taking-...
huge thanks to everyone that participated in person & remote
for more info on what a "BoF" is & what are plans at the IETF are check out: docs.bsky.app/blog/taking-...
Reposted by fig (aka:[phil])
there was a 9-min disruption to constellation, ufos-api, and spacedust this morning due to a datacentre (fig's house) power flicker.
by 8:30am EDT, auto-recovery worked for all services, which fully recovered on their own.
by 8:30am EDT, auto-recovery worked for all services, which fully recovered on their own.
November 5, 2025 at 2:39 PM
there was a 9-min disruption to constellation, ufos-api, and spacedust this morning due to a datacentre (fig's house) power flicker.
by 8:30am EDT, auto-recovery worked for all services, which fully recovered on their own.
by 8:30am EDT, auto-recovery worked for all services, which fully recovered on their own.
Reposted by fig (aka:[phil])
AppViewLite changelog:
- Hide this reply
- Reduced memory consumption / GC
- Set language when posting
- Fixed muting of Reddit users
- Fixed ARM support
- Profiler: show network requests
- Fix occasional crashes when scrolling Following
- Preserve page location when switching profiles
- Hide this reply
- Reduced memory consumption / GC
- Set language when posting
- Fixed muting of Reddit users
- Fixed ARM support
- Profiler: show network requests
- Fix occasional crashes when scrolling Following
- Preserve page location when switching profiles
November 4, 2025 at 8:14 PM
AppViewLite changelog:
- Hide this reply
- Reduced memory consumption / GC
- Set language when posting
- Fixed muting of Reddit users
- Fixed ARM support
- Profiler: show network requests
- Fix occasional crashes when scrolling Following
- Preserve page location when switching profiles
- Hide this reply
- Reduced memory consumption / GC
- Set language when posting
- Fixed muting of Reddit users
- Fixed ARM support
- Profiler: show network requests
- Fix occasional crashes when scrolling Following
- Preserve page location when switching profiles
Reposted by fig (aka:[phil])
A reminder that @chadkoh.com did a step-by-step write up of how you can join the ATProto #IETF124 BoF session at 9:30am EST this morning remotely. No cost registration.
November 5, 2025 at 12:19 PM
A reminder that @chadkoh.com did a step-by-step write up of how you can join the ATProto #IETF124 BoF session at 9:30am EST this morning remotely. No cost registration.
hmm something must be happening rn
November 5, 2025 at 2:50 AM
hmm something must be happening rn
Reposted by fig (aka:[phil])
ufos.snek.cc microcosm ufos instance online 🟢
November 4, 2025 at 2:37 PM
ufos.snek.cc microcosm ufos instance online 🟢
Reposted by fig (aka:[phil])
death by a million geese bites
third installment of the @nagare.pds.bsky.ln4.net saga: ranking models, re-rankers, and the future of the project
#atproto
third installment of the @nagare.pds.bsky.ln4.net saga: ranking models, re-rankers, and the future of the project
#atproto
death by a million geese bites
in which one rando on the internet designs a general purpose feed ![million_geese.png] %at=2025-11-03T00:50:30.882Z #author_luna #recommendation_algorithms #bluesky this is an article about nagare...
l4.pm
November 3, 2025 at 4:36 AM
death by a million geese bites
third installment of the @nagare.pds.bsky.ln4.net saga: ranking models, re-rankers, and the future of the project
#atproto
third installment of the @nagare.pds.bsky.ln4.net saga: ranking models, re-rankers, and the future of the project
#atproto
Reposted by fig (aka:[phil])
a quick 30 minute feature anyone ????
November 3, 2025 at 12:12 AM
a quick 30 minute feature anyone ????
Reposted by fig (aka:[phil])
you can use wafrn bites on reddwarf.app
November 3, 2025 at 12:50 AM
you can use wafrn bites on reddwarf.app
Reposted by fig (aka:[phil])
At #IETF124 hackathon, Table 27 with the #ATProto_MTL crew @wesleyfinck.org @ngerakines.me @bnewbold.net @bad-example.com @dholms.xyz
November 2, 2025 at 3:36 PM
At #IETF124 hackathon, Table 27 with the #ATProto_MTL crew @wesleyfinck.org @ngerakines.me @bnewbold.net @bad-example.com @dholms.xyz
Reposted by fig (aka:[phil])
couple ways to answer that!
there are some AT-specific rules, like usage of $type, and best practices like record data size limits, "nesting" depth. those are not CBOR-specific.
for encoding rules, I think we want DASL, which aligns with DAG-CBOR...
there are some AT-specific rules, like usage of $type, and best practices like record data size limits, "nesting" depth. those are not CBOR-specific.
for encoding rules, I think we want DASL, which aligns with DAG-CBOR...
November 2, 2025 at 12:33 AM
couple ways to answer that!
there are some AT-specific rules, like usage of $type, and best practices like record data size limits, "nesting" depth. those are not CBOR-specific.
for encoding rules, I think we want DASL, which aligns with DAG-CBOR...
there are some AT-specific rules, like usage of $type, and best practices like record data size limits, "nesting" depth. those are not CBOR-specific.
for encoding rules, I think we want DASL, which aligns with DAG-CBOR...
repo spec question: what's the future of record encoding, DAG-CBOR or just CBOR?
ietf draft spec, 2.7 Deterministic CBOR Encoding: just CBOR
atproto.com repository and data model specs: DAG-CBOR
in practice rn it's not strict DAG-CBOR (many floats in real records etc)
@bnewbold.net @dholms.xyz
ietf draft spec, 2.7 Deterministic CBOR Encoding: just CBOR
atproto.com repository and data model specs: DAG-CBOR
in practice rn it's not strict DAG-CBOR (many floats in real records etc)
@bnewbold.net @dholms.xyz
Authenticated Transfer Repository and Synchronization
This document specifies the repository and synchronization semantics for Authenticated Transfer (AT), a protocol for cryptographically-verifiable storage and distribution of structured user-controlled...
www.ietf.org
November 2, 2025 at 12:18 AM
repo spec question: what's the future of record encoding, DAG-CBOR or just CBOR?
ietf draft spec, 2.7 Deterministic CBOR Encoding: just CBOR
atproto.com repository and data model specs: DAG-CBOR
in practice rn it's not strict DAG-CBOR (many floats in real records etc)
@bnewbold.net @dholms.xyz
ietf draft spec, 2.7 Deterministic CBOR Encoding: just CBOR
atproto.com repository and data model specs: DAG-CBOR
in practice rn it's not strict DAG-CBOR (many floats in real records etc)
@bnewbold.net @dholms.xyz
Reposted by fig (aka:[phil])
october has ended! hacktober is over! but good-first-issues never sleep...
should this bot stay alive and keep posting good-first-issue tags? any requests or changes you'd like?
should this bot stay alive and keep posting good-first-issue tags? any requests or changes you'd like?
November 1, 2025 at 11:17 PM
october has ended! hacktober is over! but good-first-issues never sleep...
should this bot stay alive and keep posting good-first-issue tags? any requests or changes you'd like?
should this bot stay alive and keep posting good-first-issue tags? any requests or changes you'd like?
i have never seen mentions in bsky posts outpace *likes* globally. spam wave or somehow organic?
this is a log scale: normally bsky likes are 10x(!) more frequent than mentions in posts: this peak just hit >200x the normal baseline mention rate
this is a log scale: normally bsky likes are 10x(!) more frequent than mentions in posts: this peak just hit >200x the normal baseline mention rate
November 1, 2025 at 9:56 PM
i have never seen mentions in bsky posts outpace *likes* globally. spam wave or somehow organic?
this is a log scale: normally bsky likes are 10x(!) more frequent than mentions in posts: this peak just hit >200x the normal baseline mention rate
this is a log scale: normally bsky likes are 10x(!) more frequent than mentions in posts: this peak just hit >200x the normal baseline mention rate
Reposted by fig (aka:[phil])
foobar-repo-2 was an incredible sequel
November 1, 2025 at 9:42 PM
foobar-repo-2 was an incredible sequel
stumbling upon mildly interesting linked data in atproto
@oppi.li added @jay.bsky.team as a collaborator five times to three repos in @tangled.org records:
cassiopeia:6789/xrpc/blue.mi...
repos:
- core-fork (localhost:5555)
- test-fetch (localhost:5555)
- foobar-repo-2 (localhost:6000)
@oppi.li added @jay.bsky.team as a collaborator five times to three repos in @tangled.org records:
cassiopeia:6789/xrpc/blue.mi...
repos:
- core-fork (localhost:5555)
- test-fetch (localhost:5555)
- foobar-repo-2 (localhost:6000)
November 1, 2025 at 9:41 PM
stumbling upon mildly interesting linked data in atproto
@oppi.li added @jay.bsky.team as a collaborator five times to three repos in @tangled.org records:
cassiopeia:6789/xrpc/blue.mi...
repos:
- core-fork (localhost:5555)
- test-fetch (localhost:5555)
- foobar-repo-2 (localhost:6000)
@oppi.li added @jay.bsky.team as a collaborator five times to three repos in @tangled.org records:
cassiopeia:6789/xrpc/blue.mi...
repos:
- core-fork (localhost:5555)
- test-fetch (localhost:5555)
- foobar-repo-2 (localhost:6000)
summary of the small Montreal relay and jetstream incident this morning from a suspected networking issue:
- relay PDS connections dropped around 3am EST
- autoreconnect brought them back up around 8am EST
firehose event delivery was *delayed* during those five hours (no events were lost)
- relay PDS connections dropped around 3am EST
- autoreconnect brought them back up around 8am EST
firehose event delivery was *delayed* during those five hours (no events were lost)
follow-up: it's a weird one. relay logs are out of retention unfortunately so i'm not sure if i'll get to a full root-cause.
the server did not reboot, the relay container stayed up. dmesg on the host has nothing since oct 25
pds hosts connected dropped abruptly and there are some gaps in metrics
the server did not reboot, the relay container stayed up. dmesg on the host has nothing since oct 25
pds hosts connected dropped abruptly and there are some gaps in metrics
November 1, 2025 at 9:26 PM
summary of the small Montreal relay and jetstream incident this morning from a suspected networking issue:
- relay PDS connections dropped around 3am EST
- autoreconnect brought them back up around 8am EST
firehose event delivery was *delayed* during those five hours (no events were lost)
- relay PDS connections dropped around 3am EST
- autoreconnect brought them back up around 8am EST
firehose event delivery was *delayed* during those five hours (no events were lost)