VitorPamplona
Vitor Pamplona

Nostr's Chief Android Officer - Amethyst Social

Public Key

npub1gcxzte5zlkncx26j68ez60fzkvtkm9e0vrwdcvsjakxf9mu9qewqlfnj5z

NIP-05 Address

Profile Code

nprofile1qqsyvrp9u6p0mfur9dfdru3d853tx9mdjuhkphxuxgfwmryja7zsvhqpzemhxue69uhhyetvv9ujuurjd9kkzmpwdejhgqgkwaehxw309amxjar0wghxummnw3erztnrdakszyrhwden5te0dehhxarj9emkjmn9qyxhwumn8ghj7mn0wvhxcmmvqy28wumn8ghj7un9d3shjtnyv9kh2uewd9hsz8thwden5te0dehhxarj94c82c3wwajkcmr0wfjx2u3wdejhgfkrr95

Publishing to

nostr-pub.wellorder.net

Author Public Key

npub1gcxzte5zlkncx26j68ez60fzkvtkm9e0vrwdcvsjakxf9mu9qewqlfnj5z

Show more details

Last Notes

2026-05-08 23:16:28 CEST

- reply

Not really. The proof file inside the ots event already carries the tx and the block merkle path. There are two merkle structures + the transaction itself sitting between the document hash and the block header's merkle root. The elegant trick of OTS is that every byte of that intermediate material is embedded inline in the .ots proof file as APPEND/PREPEND operands. The verifier never goes to the chain for any of it. So, in practice, the verifier only executes the ops from the event, ends up with 32 bytes, and does one comparison — result == header.merkleRoot. Failure of any of the embedded data (wrong sibling, wrong tx byte, wrong path) makes that final comparison fail. There's only one check but, conceptually, you're right that two merkle trees + a tx are being walked; they're just walked using bytes inside the proof.

2026-05-08 20:53:26 CEST

Forgot to mention, congrats to Nostr Veil (ring signatures), SatShoot, Fundstr and Attestr for winning yesterday. The quality of submissions was through the roof!
https://cdn.satellite.earth/2a8d4821512dc7141000a641af153ad5910843095cea38cec66679f7a89d6527.jpg

2026-05-08 20:46:12 CEST

I just realized I can rebuild our OTS validator to use a local copy of the bitcoin chain instead of online block explorers. We only need block headers! Nothing else from the chain is touched — no transactions, no UTXOs, no scripts, no witness data, no mempool. That is a ~75 MB download, 1–10 minutes wall-clock, ~2 s CPU, ~72 MB on disk directly from the Bitcoin Core network.
Looks like an easy win. Concerns?

2026-05-08 04:25:00 CEST

https://npub1gcxzte5zlkncx26j68ez60fzkvtkm9e0vrwdcvsjakxf9mu9qewqlfnj5z.blossom.band/ec0baf4ddb9297258b6eb849a5eecef0e2ef1d892b5377a83695a73d89a1225e.jpg

2026-05-06 04:52:15 CEST

Make sure you are the crazy one

2026-05-05 04:49:29 CEST

- reply

We just moved our icons to a font for better performance :)

2026-05-05 04:15:00 CEST

- reply

How many relays does it connect to? See if resetting search and index relays help.

2026-05-05 04:11:08 CEST

- reply

Just found your DM relays.. :)

2026-05-05 03:58:48 CEST

- reply

Have you setup your DM relays, otherwise I cannot answer you. But all the help is welcome. @nprofile…wpzp is leading the desktop part and we can def use UI improvements.

2026-05-04 05:02:35 CEST

- reply

But there is always the next bug

2026-05-04 04:38:11 CEST

Everything is OpenSource now.
#nevent1q…uzg0

2026-05-04 03:00:16 CEST

When you realize the DNS resolution from Linux is a hot mess of blocking code...
https://github.com/vitorpamplona/amethyst/pull/2714

2026-05-04 02:54:01 CEST

- reply

I dislike all web interfaces on Mobile. It's cute as a feature but nobody seriously use them.

2026-05-03 15:55:37 CEST

- reply

Imagine thinking it did not. Tell me more about the human ego.

2026-05-03 03:20:48 CEST

- reply

Most of the time if the file has already been in the web, it was already compressed, so our compressor tries but can't do much.. :)

2026-05-03 01:34:01 CEST

- reply

Like the follow one? Yeah, we can add

2026-05-03 00:30:33 CEST

- reply

Very weird, even in the browser this jpeg does open as an image. It downloads as if it was a binary file.

2026-05-02 03:26:31 CEST

- reply

Bold 12 zaps please!

2026-04-30 19:58:04 CEST

- reply

The lib does not store keys. So it is up to the app to do all of that. Then each language has its own ways..
Constant time was there for hardware wallets to protect again listening to the circuitry noise and producing the key from it. But since our signers run on computers/phones, there's so much more going on that a electromagnetic sniffer cannot rebuild the key. The noise is too high.

2026-04-30 05:37:43 CEST

- reply

That's nip 98. Bunker is not just a challenge.. there are quite a few paths to be implemented and the nip is confusing/ ambiguous in all of them.

2026-04-30 05:10:38 CEST

- reply

Mostly because the bunker spec is messy and every nip46 signer seems to do things a little different

2026-04-30 04:51:44 CEST

- reply

I can't do a patch. It's a brand new codebase, not just a fork. You can see the code here if you want to play with it.
https://github.com/vitorpamplona/libschnorr256k1

2026-04-30 04:49:37 CEST

https://cdn.satellite.earth/c318e7ad65e21ef0c569e34a40a15bf528519cfbb26981bf039077964d36fc33.jpg

2026-04-30 04:15:34 CEST

- reply

Not really.. Bitcoin and Nostr are building different things. They are free to see if they want. But most likely they will not.

2026-04-30 03:58:45 CEST

- reply

We are moving from create to build() then you get an event template, which extra methods to customize your event before signing. More complicated, yes. But more flexible too

2026-04-30 00:41:32 CEST

- reply

Low web of trust, pyramid is a relay that has a fixed list of authorized users. Wine requires payment and nostrelites only allows people that the famous folks on nostr follow. I assume that is what you want. If you want more opened relays, you can add nostr.mom, damus, primal, etc. those don't filter many people.
But I can see your posts just fine. It takes a few seconds to load from these slow relays, but posts come.

2026-04-30 00:38:14 CEST

- reply

I don't want to. My improvements are for Nostr, not for Bitcoin wallets.

2026-04-29 13:29:56 CEST

- reply

Sesame oil

2026-04-29 13:12:21 CEST

- reply

Lol. We came first. Ask them to change their logo 😅

2026-04-29 13:09:35 CEST

- reply

No, I found something that everybody said was peak human talent, studied and refined for years by the best of the best, impossible to do better.
It took AI just a few hours to shatter human egos.

2026-04-29 06:02:31 CEST

- reply

Feel free to ask for things or send stuff in a PR that can make things easier for you . Quartz was built from Amethyst (which is a complicated architecture), but we want to make it simpler.

2026-04-29 05:35:49 CEST

- reply

Yeah, Quartz is quite low level. No linked cache or database. You need to make subs by yourself. But there is a SQLlite event store inside of quartz if you want to save things.

2026-04-29 05:32:54 CEST

- reply

Here is a very simple Twitter like interface:
https://claude.ai/public/artifacts/a1b4db2c-9648-4e9f-a3e7-900104f1eb9a

2026-04-29 05:21:15 CEST

- reply

See if the first sections here help
https://github.com/vitorpamplona/amethyst/blob/main/quartz/CLIENT.md

2026-04-29 05:17:45 CEST

- reply

We are also lost :)
What are you tryin to do?

2026-04-29 04:56:29 CEST

- reply

Yep

2026-04-29 04:44:49 CEST

- reply

Sesame oil extraction

2026-04-29 04:20:32 CEST

https://cdn.satellite.earth/548a090906196a9814e323c70c1c206b2e431f106b845292a3acddda24010645.mp4

2026-04-29 02:49:57 CEST

- reply

Around 2018 or so...

2026-04-29 01:01:35 CEST

If you want to scrutinize the power of the vibe.. the code is quite small.
https://github.com/vitorpamplona/libschnorr256k1

2026-04-28 23:32:53 CEST

- reply

Not after I added all the test cases I could find in all Bitcoin, secp, papers and nostr clients to it. :)

2026-04-28 23:32:06 CEST

- reply

Removed constant-time techniques, optimized for arm64 instructions, changed 5-52 limbs to 4-64 limb math, implemented batch verify (8x faster verify for 200 events), cached keys to avoid recomputing them, reduced memory creation, and deleted most of the lib to just keep the things that Nostr uses.

2026-04-28 22:56:50 CEST

- reply

More precisely, the strip.

2026-04-28 22:56:32 CEST

- reply

I don't hate the conference, I just hate Vegas. :)

2026-04-28 22:55:42 CEST

- reply

Give me 5 years.

2026-04-28 22:49:18 CEST

Soon on Amethyst
#nevent1q…sjkg

2026-04-28 22:39:46 CEST

When your vibe coded libsecp Schnoor beats Bitcoin core's code even in the highly optimized x86 build. 🚀
https://cdn.satellite.earth/6d371e5b22ee7f5301ebe03ec01b5b08e1878a845628d3b23e6356b3bdd35c6d.png

2026-04-28 21:56:46 CEST

- reply

This is an inbox issue. Only one of your followers was allowed to post they contact list to your inbox relays.
But also, we are phasing out the follower count. We will replace it by a web of trust follower count via nip 85

2026-04-28 21:55:22 CEST

- reply

Outbox seems fine. The inbox is a bigger issue. Only nos.lol is receiving replies, likes, zaps to your posts. The other inbox relays either charge the sender or block due to low wot.
We just need to move your posts from say last week to now to your inbox relays. When you run the sync, can you select just the past week to sync and see what it does?

2026-04-28 20:26:36 CEST

- reply

Now that you changed relays, try to see if the Relay Sync helps.. it's going to send the right events to the right relays from anywhere out there

2026-04-28 20:24:56 CEST

- reply

On your profile or on the home feed?

2026-04-28 19:52:31 CEST

- reply

So, do you see posts normally but not notifications? Because your public inbox relays are only 2, 1 is paid to post (which I paid) and the other has a filter by web of trust, so only very reputable keys can reply to your posts.

2026-04-28 19:43:17 CEST

- reply

Never. They need to add NWC... We are not going to fall for the vendor lock in mess

2026-04-28 19:42:18 CEST

- reply

We use the relay hints in all events where these pubkeys were seen. Massive DB in memory to know where things are.

2026-04-28 19:41:09 CEST

- reply

Is amethyst not connecting to your relays? How many relays is it connecting to on the left bar?

2026-04-28 04:56:03 CEST

https://cdn.satellite.earth/7c10731e239973e2663eb12de090652e96178c1e70b7af88044167b9c9bf0ed9.jpg

2026-04-28 04:35:30 CEST

- reply

Nothing like connecting to 300 -1000 relays to find stuff. :)

2026-04-27 14:56:40 CEST

- reply

Yes, blossom.band is there.. but since the name is that, nobody thinks it's nostr.build most people think it is the defunct nostr.band

2026-04-27 03:48:34 CEST

- reply

Yep,.matching the setting in each OS.

2026-04-26 04:19:16 CEST

It still surprises me how every day I find a new cool client that I have not seen before...

2026-04-26 04:18:03 CEST

👀
#nevent1q…cdat

2026-04-26 04:11:37 CEST

- reply

It's addictive isn't it?

2026-04-26 04:08:55 CEST

- reply

I can export a multiplatform lib once we are done.. hopefully that can help some AIs to push this into other clients. The goal is to make nests a Blossom like spec.

2026-04-26 03:48:45 CEST

lol. We are coding QUIC from scratch because no library does what we need to do. Native NostrNests Media over Quick incoming!
https://github.com/vitorpamplona/amethyst/pull/2577

2026-04-25 22:12:06 CEST

- reply

That's why we render as an image first

2026-04-25 18:28:58 CEST

👀
#nevent1q…hycg

2026-04-25 18:12:53 CEST

- reply

I like your interactive mode. We will see how far we can take Amy.. it started as a simple way for AI to test Amethyst without having to run a full app, but maybe other people can use too.. who knows...

2026-04-25 18:09:48 CEST

- reply

Fix coming this Monday

2026-04-25 18:09:39 CEST

- reply

Yeah. It's a bug on Arti Tor that saves the state in disk, so even if you kill the app, it won't connect anymore :(

2026-04-25 13:46:02 CEST

- reply

Fancy! Did it work well with addressables and expiration?

2026-04-25 06:51:51 CEST

Remember this? We actually coded it and it works! A full nostr event store that saves each nostr event as a visible prettified json file. You can browse your db, delete files, r sync it, everything will stay working. We even have full text search with file links 😅
#nevent1q…ka7y
https://github.com/vitorpamplona/amethyst/pull/2560

2026-04-25 04:37:12 CEST

- reply

See if turning airplane mode on and off fixes it

2026-04-25 04:36:40 CEST

- reply

Shooting for a new release on Monday that should fix this

2026-04-25 04:36:10 CEST

👀👀👀 https://cdn.satellite.earth/13649504c19cb3ef3b6f34356a19c105fb1483d71e90edaac699dc333d90968e.mp4

2026-04-25 01:27:49 CEST

- reply

I was thinking on moving to your style of icon + text in horizontal and add the chat right there too. So, everything becomes a double pane design...
Because we will put a lot more things in that menu.. just the different relay feeds alone will be huge.

2026-04-25 00:26:54 CEST

- reply

It is still quite rough, but we are getting there.

2026-04-24 18:41:01 CEST

- reply

@nprofile…prwr we could use your help in the UI design

2026-04-24 18:37:32 CEST

Amy Desktop is getting a custom design for each OS, including MacOS, Gnome, KDE, and even Windows. Here's MacOS:
https://image.nostr.build/df2f29b02ed75e6abf35f158c5eea0557b67d54b922c2f795a2bce74b86f9d31.png
https://image.nostr.build/c7806e837d7ae2b256e56f1378becd8674c4fe3b37dec6f915e7c5eba5e43c30.png

2026-04-24 18:17:57 CEST

- reply

I also don't rewrite... I write it better. WAY better.

2026-04-24 14:39:27 CEST

- reply

Performance improvements are coming... Our struggle has always been the balancing act of preloading a lot of stuff so that when you switch to any tab all the info is already there and the budget we have (500mb) in older phones. I think we do well in newer phones because the budget is 1GB.
Sometimes even rendering icons and clickable sections in the post can be slow. 🫤

2026-04-24 04:23:45 CEST

- reply

Yep. By FAR.

2026-04-24 00:28:40 CEST

- reply

I have 15 Claude work streams in parallel. All I do is to swap between them, test and get them to fix it

2026-04-23 18:59:03 CEST

Amethyst saves
#nevent1q…aytv

2026-04-23 18:55:14 CEST

- reply

These days, yes. Testing is the bottleneck

2026-04-23 18:54:59 CEST

- reply

It is... I am surprised I am still here doing this.

2026-04-23 18:54:40 CEST

- reply

I need to use a bunch of extensions to load nostr info about pages or things I am seeing... They all have too many ads, even those that remove ads from the page track us for their own ads. They all consume 10x more memory per page than what I would like. They are All to bloated or heavy because they support 40 years of web standards. Chromium monoculture must die. Safari needs to die as well.

2026-04-23 17:15:15 CEST

- reply

Yep, we do that too first

2026-04-23 17:11:14 CEST

My day-to-day in the age of AI on Nostr
https://image.nostr.build/7a5b96b01dd1ffc2466249673ab5dd798f209f1db85acb64a4742e1a8294cfcb.png

2026-04-23 13:50:08 CEST

- reply

That too

2026-04-23 13:30:09 CEST

- reply

Sure, but all browsers suck right now, so I would welcome a brand new one.

2026-04-23 13:29:14 CEST