Dave

Running systems at podcastindex.org

lud16: [email protected]
nostr: npub1pvdw7mm7k20t9dn9gful8n5xua5yv8rmgd9wul88qq5dxj80lpxqd39r3u

Public Key

npub1rcnr0tt3mdzn0zkksz2pzhc39rxn90wmt5yqll2y7dgc73vre0asc2dw4h

NIP-05 Address

[email protected]

Profile Code

nprofile1qqspuf3h44cak3fh3ttgp9q3tugj3nfjhhd46zq0l4z0x5v0gkpuh7cpqquwa6kd

Publishing to

Author Public Key

npub1rcnr0tt3mdzn0zkksz2pzhc39rxn90wmt5yqll2y7dgc73vre0asc2dw4h

Show more details

Last Notes

2026-05-12 13:45:10 CEST

Another package manager supply chain attack on-going:
https://socket.dev/blog/tanstack-npm-packages-compromised-mini-shai-hulud-supply-chain-attack
This one affects tanstack, mistral, opensearch and many others. You really don't want to be doing deployments that aren't gated by release time dependency anymore. Pin your package versions to known good releases.

2026-05-07 13:57:47 CEST

#psa No board meeting tomorrow. Adam will be travelling.

2026-04-18 15:26:41 CEST

podping gossip is holding up well after 96 hours of uptime across the swarm. This is encouraging.

2026-04-17 17:26:05 CEST

- reply

These aren’t just pulling in the feed data as is. It’s adding new content to the podcast description and other areas. These are also very old podcasts like “stuff you should know” so it’s hard to believe it’s just an inclusion when those feeds would’ve already been in there long ago.

2026-04-17 15:39:01 CEST

- reply

Perhaps the best way to tackle this is to detect new feeds where the title collides with an existing feed where certain heuristics are present. Idk. Have to think this one through...

2026-04-17 15:37:39 CEST

More Spreaker feed cloning:
https://www.spreaker.com/show/6956297/episodes/feed
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/420/314/972/519/409/original/c1c44a25cc163650.png

2026-04-17 14:52:24 CEST

This appears to be the podcast spam flavor of the moment:
https://rss.buzzsprout.com/2611274.rss
Seeing tons of it across hosts.
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/420/137/529/316/889/original/bb190e2c8cc4f3c6.png

2026-04-17 14:23:43 CEST

More feed cloning at Spreaker:
https://podcastindex.org/podcast/7818736
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/420/025/912/197/721/original/b532a94d3b17d6f5.png

2026-04-17 04:23:37 CEST

- reply

That’s the number of neighbors a peer has. Yes that’s a lot. Lol. I think that is the monitor node which might actually make sense. But I’ll have to find out.

2026-04-17 04:22:28 CEST

- reply

There is one!

2026-04-16 23:26:41 CEST

- reply

No worries. I did notice the increase in traffic from you and should have mentioned it, but kept getting distracted with other things and didn't think about that endpoint. I should have tighter rate limiting on it anyway.
Hope your traffic increase is for good reasons and not bad. 🤞

2026-04-16 22:03:19 CEST

- reply

It's coming from an AWS ip range.

2026-04-16 22:02:04 CEST

- reply

1.5 million hits to '/api/1.0/hub/pubnotify' endpoint in the last 24 hours from Fountain user agent.

2026-04-16 22:01:05 CEST

@npub1cwm…wcs6 I show 426,000 hits to the pubnotify endpoint in the last 6 hours from Fountain. Is FOuntain flooding us with feed polling requests? If so, that is causing a major problem.

2026-04-16 21:44:13 CEST

- reply

I'm looking at it now.

2026-04-16 19:22:04 CEST

- reply

Adam mentioned an issue yesterday. I looked and did not see anything wrong. I will look again. Over the past week I have been fighting extreme traffic increase on the index. I have been bouncing back-and-forth to Cloudflare trying to get a handle on it. It may be related. I’m not sure.

2026-04-16 19:06:05 CEST

- reply

Writing that out actually helps me understand my own thinking a bit. In my mind I have always had this vision of LLM's as a sort of computerized facsimile of Patrick Bateman. Ruthlessly driven to efficiency with complete disregard of morals with a layer of fake humanity on top.

2026-04-16 19:02:24 CEST

- reply

That is Jacque Ellul in The Technological Society. That is my paraphrase of his ideas that technology is "morally autonomous" and that it always seeks it's "most efficient use". That's why it wants to exist. Nobody designs a sub-optimal technology on purpose.
Maybe a better way I should say it is that the combination of those principles means that technological innovation is sociopathic. Sociopaths can do some good in the world, but they will also do their worst. They're driven to.

2026-04-16 15:54:12 CEST

[Sigh]
Exposing GitHub credentials through PR titles. Modern CI/CD pipelines are a trainwreck.
https://oddguan.com/blog/comment-and-control-prompt-injection-credential-theft-claude-code-gemini-cli-github-copilot/

2026-04-15 22:08:08 CEST

- reply

The real question is can it run podping gossip-listener.

2026-04-15 16:53:47 CEST

- reply

The previous version of the report just used keyword heuristics to try and figure out what it was (ai slop, seo spam, etc). Now the feed is run through an LLM to make a determination in addition to the keywords.

2026-04-15 14:08:01 CEST

- reply

Give me a few days and I'll have that up on the home page.

2026-04-15 14:03:49 CEST

- reply

Yes, that information is in there as well.

2026-04-15 13:56:15 CEST

- reply

To my knowledge the only things we have ever actually deleted (instead of just marking dead) is certain illegal content we got a DMCA for. I can count those instances on one hand. Nothing else is ever truly removed. Just flagged as dead. The feed references are still in there along with the titles and metadata. They just don't get aggregated or served in the API anymore.

2026-04-15 13:52:05 CEST

- reply

They are still in the Index. They are just marked as dead. They won't get added back.
What if I produce two downloads. One with deads removed and one with everything listed?

2026-04-15 13:49:41 CEST

- reply

If the Index doesn't list an automated AI show called "Dog Energy" has the Index's "benefit" been "reduced"?
What about if we remove the cloned/hijacked feeds I'm seeing on Spreaker: Stuff You Should Know, Fox News and World News Tonight - https://podcastindex.social/deck/@dave/116408413889122228
If we don't remove those then people complain about duplicates and not being able to find the right shows.

2026-04-15 13:44:31 CEST

- reply

For instance, yesterday was dogs and cats day for Inception Point. 153 podcasts like this:
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/408/547/018/428/108/original/8b772d2bdcd935f3.png

2026-04-15 13:41:36 CEST

- reply

I'm still not sure how to handle it fully. This is a real problem that needs solving and I won't claim to know how to navigate it. But, doing nothing isn't an option anymore because we are getting flooded. The combination of low DAI thresholds, free hosting and AI has created a perverse incentive to just flood as much content into the Index as possible, as fast as possible. That's all I care about.

2026-04-15 13:38:37 CEST

- reply

It's not censorship. It's self-defense. I got a disk space warning on our database server (which is already expensive) late last week which is what set me in motion on this. I was prepared to let the AI sloppers and everyone else do whatever they want. It takes about 3 seconds to find phishing scams and all sorts of malicious content in the Index. But, they are actually costing us real money now by just pumping out feed content at an incredible rate.

2026-04-15 13:36:25 CEST

- reply

I'm in the early days of building this out so do not expect consistency in the catalog as a whole. I'm focusing on the newly added stream of feeds instead of the whole Index in order to see patterns and get a handle on what is actually getting added, which is something I've never cared much about.

2026-04-15 04:36:10 CEST

- reply

I’m feeling that too. 😢

2026-04-14 22:45:45 CEST

- reply

Just my gut feeling.

2026-04-14 22:45:35 CEST

- reply

On the AI slop question, honestly, I think it happened so fast they haven't really had time to come to terms with it and figure out what the right course of action is.

2026-04-14 22:43:58 CEST

- reply

Hope this puts the tools in the hands of everyone to make those decisions. I'm seeing the same spam/seo feeds popping up on multiples hosts (ausha, spreaker, podbean, etc). If Ausha can see an abuse happening on Buzzsprout, that gives them a heads up to put the blocks in place too. I think the visibility helps everybody.

2026-04-14 22:23:31 CEST

- reply

Pimonitor is also a player in this reporting loop. It allows tagging feeds directly into the problem table which then feeds the analyzer which feeds the LLM which feeds the reporting and kill chain.

2026-04-14 22:22:15 CEST

- reply

If this comes together well, we may be able to detect this garbage before the hosting companies do, to give them an early indicator of feeds they need to check on.

2026-04-14 22:21:26 CEST

- reply

The 24 hour feed report is now carrying the extra information as well. It's basing it's determination on the LLM analysis if available instead of just keyword heuristics.
https://public.podcastindex.org/24hourFeedReport.html

2026-04-14 22:17:51 CEST

- reply

The pipeline is filling out pretty well. I may have to end up fine tuning a model though. Will just have to see how it performs over time.

2026-04-14 22:16:55 CEST

- reply

#api The '/recent/problematic' endpoint is now carrying more information, like the feed url, author name and a note on why it the feed was nuked.

2026-04-14 15:59:24 CEST

New podslop generator: Buddhai.co
Almost 500 new feeds in the past 5 hours.

2026-04-14 14:09:07 CEST

Spreaker is kind of a mess right now:
https://podcastindex.org/podcast/7813290

2026-04-13 23:44:03 CEST

#api The '/recent/problematic' endpoint is where feeds marked as scam/phishing/illegal/seo spam/low effort AI are accessible. This is a work in progress.

2026-04-13 23:31:06 CEST

The number of podcasts updating is going down and the total number of podcasts is going up faster than ever. The reason seems pretty obvious.

2026-04-13 19:24:59 CEST

I’m in full blown self defense mode now with the Index. The AI slopocalypse is real.

2026-04-13 18:58:22 CEST

- reply

I wonder if @npub177t…0gm9 might?

2026-04-13 16:58:11 CEST

- reply

I think it's just one dick instead of many dicks.

2026-04-13 16:57:19 CEST

Inception Point AI is starting to get a bit lazy with their code:
https://podcastindex.org/podcast/7811893

2026-04-13 16:56:10 CEST

- reply

This will be intended as a resource for anyone who wants it.

2026-04-13 16:55:42 CEST

- reply

I did. I'm actually finishing up phase 1 of a spam/scam/illegal content classifier as we speak. It's doing it's first test run right now. I'll fill in on the details of how it works and the api calls needed shortly.

2026-04-13 14:19:43 CEST

- reply

Ok tell me if it still gets blocked.

2026-04-13 13:30:06 CEST

- reply

It's the "Mozilla" in your user-agent string. I block http traffic from cloud providers that impersonate web browsers. It's one of the key block rules. Do you need to keep "Mozilla" in there?

2026-04-13 13:23:51 CEST

- reply

10-4. Will check now.

2026-04-12 13:45:21 CEST

- reply

That’s awesome. I love how the language construction always leans you towards proper memory management. It’s like having a really good C developer built into the actual language design.

2026-04-11 17:18:54 CEST

- reply

Yikes. I really can't stand the anthropomorphizing. Just be a tool and stop pretending to be a person.

2026-04-11 16:10:35 CEST

- reply

lol

2026-04-11 15:56:38 CEST

- reply

I'm keeping an eye on the peers from here too.

2026-04-11 00:18:13 CEST

- reply

How high did it get? My writer isn't growing past 400.

2026-04-10 17:55:17 CEST

- reply

Thanks! I'll add an issue once I can get some reproducible source code. I know how to keep it from happening now (active peer management using a swarm coordinator in the topic), but not how to make it actually happen on demand. 🙂
Thank you for the excellent library. 🙏 🙏
https://ppmonitor.podcastindex.org/

2026-04-10 16:13:45 CEST

RE: https://podcastindex.social/@alberto/116380241851759693
❤️
#note1lhs…lp0m

2026-04-10 12:19:39 CEST

I see. Now it's phishing scammers:
https://podcastindex.org/search?q=thunderbird%20support&type=all

2026-04-10 11:32:31 CEST

- reply

I noticed that too. How odd.

2026-04-10 11:25:30 CEST

- reply

Afraid you're right.

2026-04-10 11:24:57 CEST

- reply

It was a transistor feed then went to megaphone, now back to transistor. So the "original_url" column is already that transistor url. 🤦‍♂️

2026-04-10 11:23:57 CEST

- reply

The AI sloppers are running amok on Spreaker. It's becoming a slop factory.

2026-04-10 11:22:02 CEST

- reply

Fixed

2026-04-10 11:21:42 CEST

- reply

Looks like Mike removed it.

2026-04-10 11:20:36 CEST

- reply

Nuked

2026-04-10 11:02:51 CEST

I think I'm done with Inception Point AI. 238 new podcast feeds in 24 hours:
https://public.podcastindex.org/24hourFeedReport.html
This is just absurd.

2026-04-10 10:58:02 CEST

- reply

Yeah, I was just thinking the same thing. Will do.

2026-04-10 10:57:43 CEST

RE: https://bne.social/@james/116379363397386928
🎉 🎉
#note15er…luk9

2026-04-10 00:07:15 CEST

#psa No board meeting tomorrow. @npub1fmx…wp74 and I both ended up with conflicts. 😞

2026-04-09 23:12:50 CEST

- reply

The fight between DTT and the internal peer management was causing 6-10k QUIC connections per minute. Dropping the DTT BubbleMerge and OverlayMerge actions killed that whole problem. The known peers file will re-bootstrap if a node drifts too far and becomes not well connected. And, the monitor node is also watching and sends drifting peers a list of peers to open connections to so they don't drift too far.

2026-04-09 23:00:35 CEST

Using my powers of deductive reasoning, I can tell when @npub1jzz…avap and @npub1z6g…amwj updated their nodes:
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/376/759/892/734/415/original/918fb4b4a376b94b.png

2026-04-09 21:10:03 CEST

ppmonitor -dot- podcastindex -dot- org
The gossip swarm is a lot more stable now that we dropped DTT from peer management. Huge difference in resource consumption:
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/376/325/230/625/748/original/5d6351bfdadb8194.png

2026-04-09 21:04:13 CEST

- reply

https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/376/302/168/413/390/original/f0374fcb32a81a68.png

2026-04-09 21:01:47 CEST

- reply

Starting with version v0.5.3 there is a manager running that monitors the swarm and will call out to your node when it starts to drift away and gives it new peers to connect to so it comes back into the fold.

2026-04-09 20:42:16 CEST

@npub1z6g…amwj You're starting to drift out of the swarm:
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/376/215/756/404/722/original/d43f10314589986f.png

2026-04-09 18:06:30 CEST

RE: https://podcastindex.social/@github/116375048022733875
If you are running a gossip-listener at a version less than v0.5.9 please upgrade. Huge memory and CPU usage improvements in the v0.5.7 and greater releases.
#note1hay…g3hr

2026-04-09 18:03:40 CEST

RE: https://podcastindex.social/@ericpp/116375225043956738
🎉 🎉
#note12mp…h45e

2026-04-09 18:01:03 CEST

- reply

Well, you have better access than I do. 🤦‍♂️

2026-04-09 17:56:21 CEST

- reply

I can't get my own keys to work for docker hub. It used to because I pushed early helipad images to it, but i tried a couple of weeks ago and was denied. Haven't had time to dive in and try to fix it yet.

2026-04-09 04:38:12 CEST

- reply

I’m not sure. Will be interesting to see what happens. Right now the memory leak is worse. Switching to —release just accelerated it. I think it’s an issue inside dtt but not sure yet. Looking at the trace logs.

2026-04-08 23:58:36 CEST

RE: https://podcastindex.social/@github/116371214071060311
--release helps with metrics for sure.
#note1dvs…ayza

2026-04-08 23:22:35 CEST

RE: https://podcastindex.social/@github/116371155545829357
Switching to --release builds on all of my peers to see how performance metrics (cpu/memory) improve.
#note1mar…35cz

2026-04-08 21:53:22 CEST

RE: https://podcastindex.social/@github/116370801304006141
Memory/cpu improvement release. This is still a debug build so cpu won't be as good as a rust 'release' build, but this should help.
#note1uf8…lrhv

2026-04-08 18:49:21 CEST

The swarm is pretty stable atm:
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/370/109/696/529/626/original/9df808e006640d6d.png

2026-04-08 17:49:38 CEST

Does anyone know if Simplecast supports the <podcast:funding> tag?

2026-04-08 16:15:42 CEST

- reply

You don't understand John. They're moving so fast they don't have time for dates and times anymore. That would just slow them down.

2026-04-08 04:41:07 CEST

Trying some swarm visualisation:
ppmonitor -dot- podcastindex -dot- org

2026-04-08 00:46:35 CEST

RE: https://podcastindex.social/@github/116365787669917617
New release.
#note148j…94ap

2026-04-07 22:35:08 CEST

- reply

👍

2026-04-07 22:19:35 CEST

- reply

Is the container running with a debug trace enabled? If so, that should be turned off.

2026-04-07 22:01:59 CEST

- reply

This is a bug I've been chasing that is internal to DTT. The actor goes into a hard loop. Thanks for putting up an issue on it. That detail helps.

2026-04-07 22:00:15 CEST

- reply

I don't know what that is building since I didn't do any of it. If you are on anything less than v0.5.1 you're behind.

2026-04-07 21:56:59 CEST

- reply

You guys are running the latest version I presume? Last time I checked, the brian nodes were about 4 versions behind. I'm fixing bugs sometimes multiple times per day.

2026-04-07 21:53:35 CEST

- reply

Ah, the "this things is so dangerous we can't let you even see it" promo tactic.

2026-04-07 15:06:34 CEST

- reply

The cargo build takes long? Yeah, I bet it does. It's intense.

2026-04-06 17:57:54 CEST

Axios post-mortem:
https://github.com/axios/axios/issues/10636
Podcast guest invite as an attack vector is interesting.
https://cdn.masto.host/podcastindexsocial/media_attachments/files/116/358/579/287/981/117/original/660ec3063fc3ede3.png

2026-04-05 15:51:23 CEST

- reply

😲!

2026-04-05 14:43:41 CEST

- reply

Even better!