hodlbod

stuff.coracle.social

Christian Bitcoiner and developer of coracle.social. Learn more at info.coracle.social.

If you can't tell the difference between me and a scammer, use a nostr client with web of trust support.

Public Key

npub1jlrs53pkdfjnts29kveljul2sm0actt6n8dxrrzqcersttvcuv3qdjynqn

NIP-05 Address

hodlbod@coracle.social

Profile Code

nprofile1qqsf03c2gsmx5ef4c9zmxvlew04gdh7u94afnknp33qvv3c94kvwxgspr3mhxue69uhhyetvv9ujumt4w35ku7thv9kxcet59e3k7mgpnfmhxue69uhkummn9ekx7mp0y5erqamnwvaz7tmwdaehgu3wd3skuep0y5erqffjxpshvct5v9ez2v3swaehxw309ahx7um5wgh8w6twv5hj2v3sy5erqctkv96xzu39xgc8wumn8ghj7ur4wfcxcetjv4kxz7fwvdhk6te9xgc8wumn8ghj7un9d3shjtnyv9kh2uewd9hj7ffjxpmhxue69uhhyetvv9ujuumwdae8gtnnda3kjctv5gddy8

Publishing to

relay.mutinywallet.com

nos.lol/%20wss://nostr.land/%20%20avatar%20wss://nostr.wine/%20%20avatar%20wss://purplerelay.com/%20wss://relay.damus.io/%20wss://relay.snort.social

nos.lol

Author Public Key

npub1jlrs53pkdfjnts29kveljul2sm0actt6n8dxrrzqcersttvcuv3qdjynqn

Show more details

Last Notes

2026-03-25 21:23:33 UTC

- reply

I think for the most part it doesn't ruin motivation if you're already motivated. But it may skew incentives so that you're more likely to do interesting experiments instead of building a solid product.

2026-03-25 19:02:30 UTC

- reply

And how do you acquire taste? By doing the work.
I think it's sort of troubling how unequipped we are to apply the taste we have acquired by doing the work for decades. People who started with LLMs will have nothing to offer. Or, maybe they will acclimatize and focus on taste and craftsmanship.

2026-03-24 20:46:25 UTC

- reply

Do you use spec driven development or what? I'm currently experimenting with "writing" the app in excruciating detail in markdown (but using real function signatures and names) and having the AI fill in the details. For example:
```
## `async fn create_relay(...) -> Response`
- Serves `POST /relays/:id`
- Authorizes admin and relay owner
- Creates a new relay using `self.repo.create_relay`
- If relay is a duplicate by subdomain, return a `422` with `code=subdomain-exists`
- Return `data` is a single relay struct. Use HTTP `201`.
```

2026-03-24 17:50:18 UTC

- reply

Haha yes, that sounds like me

2026-03-24 17:50:00 UTC

- reply

I like DDD, even read a book, but I've never really had the opportunity to use it. I do think it could be a really effective tool here.

2026-03-24 17:49:26 UTC

- reply

I keep trying, and the LLMs are getting better and better at producing the illusion of a codebase. But sometimes a single line exhibits a complete lack of comprehension of anything related to what it's supposed to be doing. It's so tantalizing.

2026-03-24 17:43:58 UTC

Farming isn't the production of food, it is the progressive discovery of what food to produce on your particular land, and how to do so while balancing ecological health and economic sustainability.
#nevent1q…f3gf

2026-03-24 17:37:50 UTC

- reply

But I think that probably puts the utility of llms at a much lower level — at the level of syntax, or at least functionality. Naming functions is crucial to developing a nomenclature for a project, so LLMs can't really be useful at that level or higher (except for functions that you don't care about — in which case, why not just write a big procedural blob).

2026-03-24 17:36:09 UTC

Most code we write is a mistake. Cutting ourselves off from understanding the code we write cuts us off from understanding our mistakes, which makes it impossible for us to improve.
LLMs are not merely higher-level tools, because they are not themselves abstractions. Abstractions are refined over decades to hide the underlying complexity (and even then fail). Forget leaky abstractions, LLMs are pure leak, no pipe.
#nevent1q…4r8p

2026-03-24 17:34:02 UTC

How much code survives 10 years of an application's development? How much of that development is a result of internalization of problems solved and articulation of new mental models? Is product development a direct result of the personal growth of the individual/community (team) working on the product?
#nevent1q…4r8p

2026-03-24 17:32:27 UTC

Vibe coders claim that the mental model is disposable, what's important is the code you produce. But what if it were the opposite? That the code is disposable, what's important is the mental model you produce?

2026-03-24 16:01:01 UTC

- reply

Is there a spec?

2026-03-24 00:41:47 UTC

- reply

No, like, what are the relay urls/implemenations, monitor pubkeys, docs, and agents

2026-03-23 21:35:50 UTC

- reply

that's pretty messed up lol

2026-03-23 19:37:13 UTC

- reply

Cool, I'm glad to know that's the sort of thing you're hoping to address. Where is this stuff?

2026-03-23 18:46:01 UTC

- reply

Yeah, I agree with that. Like I said, it's useful in theory, I just haven't heard anyone describe how it's actually supposed to be used in any particular case.

2026-03-23 18:29:06 UTC

- reply

The problems I'm talking about are true of any automatic consumption and use of monitor events. That can be solved through sophisticated analysis, or through manual review. A thing like nostr.watch that would allow client developers to find and select good relays would be a welcome tool (nostr.watch's tools are too rudimentary to be super helpful). Likewise, a sophisticated server component that client devs can run that analyzes monitors and exposes relay selections via API could also be useful. My point isn't that raw monitors aren't useful, but that they need lots of opinionated tooling on top to become useful.

2026-03-23 18:20:38 UTC

- reply

No one should have been relying exclusively on nostr.band, that's the whole idea with relays. The same problem would be true of bootstrap relays for finding monitor information anyway.

2026-03-23 17:47:46 UTC

- reply

In order to "hardcode" a monitor, you need to:
- Know its pubkey
- Know which relays it publishes to
- Know you can trust it
If you run the monitor or trust the person who runs it, this is easy and helpful. It's centralized, but no more so than hardcoded relay selections. The downside is you have to do a bunch of analysis of monitor events in the client before you select relays for other purposes. You could maybe mitigate this by doing this computation server-side and caching it. But then the monitor is really just an infrastructure piece for the client and doesn't need to be interoperable or have. NIP.
A NIP is for interop, which implies monitor discovery. But in order to do that you need to:
- Have a hardcoded list of relays that host monitor events (or a hardcoded bootstrap monitor), any of which are trusted actors who can suppress or sybil attack monitor listings and by extension monitor events
- Be able to compute web of trust or other trust mechanisms for monitors, which implies a hardcoded list of relays where you can get 10002 events (or a dependency on something like vertex).
- Do all the same things related to fetching monitor events and analyzing them.
It just feels like a Rube Goldberg machine to me. Compared to hard coding relays, it has the same bootstrapping relay selection requirements with a much higher level of complexity involved in network traversal and analysis. Which again, can be fine, but isn't really feasible client-side and so ends up being as centralized as a proprietary solution based directly heuristics like user relay lists.

2026-03-23 17:32:19 UTC

- reply

Looks interesting from the example, is there documentation or anything?
https://github.com/damus-io/notedeck/tree/master/crates/notedeck_wasm

2026-03-23 16:31:05 UTC

- reply

😂 yes

2026-03-23 16:27:31 UTC

- reply

Or ads

2026-03-23 16:27:14 UTC

- reply

Yeah, because the client controls the context in which the scroll is invoked, the display function can be infinitely overloaded and presented as buttons, notes, grids, whatever.
The use case I want to solve is custom tabs in Flotilla. Basically, the relay would publish one or more self-signed scrolls with a name (or a list of scrolls), and those would show up as tabs in the navigation. Click on the tab, and you get a page based on some presentation template suggested by the list/scroll. The thing that's missing is interactivity I suppose. It would be nice for the mini app to be able to define clickable buttons, inputs, etc.

2026-03-23 16:22:38 UTC

- reply

Nostrscript is the thing @nprofile…36hd used to brag about which eventually became nostrdb (I think). AFAIK a spec or library was never published. I also played around with a lua scripting language: https://github.com/coracle-social/ostrich-script-js

2026-03-23 16:19:54 UTC

I would consider getting a neuralink if it could be configured to prevent me from seeing a My Little Pony gif ever again

2026-03-23 03:09:17 UTC

- reply

Oh, maybe you're agreeing with me, I can't tell

2026-03-23 03:08:29 UTC

- reply

This NIP seems to leave a lot of the control up to the client, with only one render specific method

2026-03-23 03:07:38 UTC

- reply

Explain what the mechanism is for validating that monitors are being honest

2026-03-22 20:15:40 UTC

- reply

Vervaeke cries 😂

2026-03-19 20:36:00 UTC

- reply

It's on shuffle, so all of them, and I haven't dug into The New Abnormal yet, but it's promising unlike most new albums from old bands

2026-03-19 19:33:57 UTC

On day 3 of just listening to The Strokes

2026-03-19 19:32:53 UTC

One thing I can say in favor of LLMs: I feel a growing sense of agency, not in terms of being able to directly do things, but in terms of being able to find out what step 1 is for doing something myself.

2026-03-19 19:31:45 UTC

- reply

might be time to switch. I've been coasting on my linux skills for a decade now

2026-03-19 19:29:10 UTC

- reply

Do you attach PoW? I'm starting to think we should have made that a requirement in the nip

2026-03-19 19:08:10 UTC

- reply

maybe we'll finally move away from systemd now

2026-03-19 17:26:14 UTC

- reply

There's no one idea of Christian politics. The gamut ranges from Monarchy to Anarchism and everything in between. I think Judges (decentralized self-governance) vs Samuel (two kings, one righteous, one unrighteous) is instructive. I think the ideal form of government is probably dependent on history, culture, geopolitics, where in the civilizational cycle you are, etc.
I do think that the stronger the government, the more important (and difficult) it is to cultivate a strong folk culture that doesn't care what the government does. This to me is Christian Anarchism — not fighting resisting the government, just doing our own thing outside of it. This is how I understand Romans 13 and 1 Peter 3: in light of Mark 12. Submit to the governing authorities, pay your taxes to Caesar, but give to God what is God's — i.e., yourself. If the government claims your allegiance, it is over-reaching. But if it claims your stuff that's ok because your treasure ought to be in heaven.
I think Christians (even Christian Anarchists) can be in positions of civil authority (although being that close to the machine can be morally hazardous). But being in authority and exercising power isn't inherently wrong; the soldiers in the New Testament are generally praised as examples of faith, and are not instructed to quit their job (Luke 3, Matthew 8, Acts 10).

2026-03-18 23:26:28 UTC

- reply

This is what true love looks like

2026-03-18 23:18:02 UTC

- reply

I just realized my math makes no sense, it's been a long week ok

2026-03-18 23:17:33 UTC

- reply

It makes them last longer, the chances of you getting the last pistachio asymptotically approach 1/n

2026-03-18 15:42:52 UTC

- reply

Because the relay *is* the group. Discord calls their groups "servers", I don't see why this should be so confusing. NIP 29 has poisoned everyone's minds by putting every group on groups.0xchat.com I guess. I want a more sophisticated type of group where admins have more control.

2026-03-18 15:41:01 UTC

- reply

This is what I'm working on today, just provide good defaults and users can set stuff up later

2026-03-18 15:40:30 UTC

- reply

In the latest version the opt-in dialog should be gone (at the cost of triggering all the expensive decryption mentioned by @nprofile…py7w ). The home page sucks and will go away soon, but you can also start a chat from the chat page. Once I fix all the stuff I already know is bad I do want your feedback, I promise.

2026-03-16 20:25:18 UTC

This is a first, my agent has corrupted my git index twice now 😬

2026-03-16 16:00:30 UTC

- reply

I've been told this metaphor isn't accurate theologically, but it's the best I have: God is like an author who chooses what his characters do, in detail and with absolute authority. But that doesn't infringe on the free will of the characters in the story, who make genuine choices within the parameters of their reality. So God's will is determinative, but life is not deterministic (i.e., our choices really matter).
The main idea is that God is both sovereign and transcendent. The mechanics of time, cause and effect, locality, etc. on which our free choices are predicated can't exist without his active involvement. At the same time, God exists outside of time and the rest of created reality, and so does not impose himself on our will, but predicates it.
Arminians pit God's will against man's. But the reality is that they are in harmony.

2026-03-15 14:58:28 UTC

- reply

Calvinism is not deterministic

2026-03-15 04:31:17 UTC

- reply

This is why there is no rioting, because we trust God, not marxist screeching. I want justice, but making a sign and standing around does not feel like the thing. I also don't know what it looks like.

2026-03-13 16:21:19 UTC

I just found the best product review ever (for a tiny 1oz steel flask):
https://hbb.coracle.social/ab8ede4c29b2ea69b2c08c3c614418744ceec1198da30cc61c41ec46d0ad46b6.png
"I have been stopped too many times." 😂😂😂

2026-03-13 12:36:30 UTC

- reply

Yep, they got acquire a year or two ago

2026-03-13 00:18:41 UTC

- reply

No, is it good?

2026-03-12 21:14:09 UTC

- reply

Thanks, I wasn't sure where to look, but that's helpful

2026-03-12 21:08:32 UTC

- reply

What is this subtweet

2026-03-12 20:58:41 UTC

@nprofile…6nzx is there a way to see what providers are charging per model/token? I have the feeling that I'm paying substantially more via routstr than via opencode/openrouter. I'm ok with some premium because it saves me having to sell my bitcoin for dollars, but I don't want to pay hundreds more per month.

2026-03-11 00:15:48 UTC

You heard it here first folks, Lopp is trans
#nevent1q…89dp

2026-03-10 01:20:21 UTC

- reply

Explain how facebook business suite is as bad as pressboard bookshelves

2026-03-09 21:35:34 UTC

- reply

> You're experiencing the same as countless other artisans have experienced in the last century.
> It doesn't matter to anyone but you.
These claims are not mutually consistent. The result of the loss of craftsmanship is disposable consumer goods and the exploitation of the environment and human life. Software already was this, but LLMs are more so. The loss of craftsmanship is the loss of everything worthwhile that was being produced, and its displacement by literal garbage.

2026-03-08 15:19:18 UTC

- reply

Awesome. My sone came and sat next to me and made this:
https://hbb.coracle.social/5564c00d280fe0209c11776814886c23b39e4bce44c0b7ff603053b4ab17f708.jpg

2026-03-08 13:41:39 UTC

I made a spoon
https://hbb.coracle.social/7b63558d4e2a008cd1029f9806a3d2b878c0551112a575d2b4a798ea6ccf3a1b.jpg

2026-03-06 22:34:52 UTC

Getting close to being able to release pomade to Flotilla users. Here's a boring video which showcases the UX:
https://coracle-media.us-southeast-1.linodeobjects.com/pomade_demo_20260306.webm

2026-03-06 00:26:34 UTC

- reply

This is very good

2026-03-06 00:05:55 UTC

- reply

Oh, I see, yes I ignore <15

2026-03-06 00:05:13 UTC

- reply

I see a 16 on this note

2026-03-05 23:10:35 UTC

- reply

Debian's days are numbered

2026-03-05 22:14:17 UTC

- reply

I don't know how true this is, but maybe because they see it as competing for funding with bitcoin, as a distraction, or even as "crypto"? I know there was some talk about this back when OpenSats started their nostr fund.

2026-03-05 20:59:21 UTC

- reply

Me too. I've used it to good effect to write some code in a domain I don't know in languages I don't know, and I can mostly verify the results by writing tests. But for domains and languages that I do know, I much prefer writing code manually (and I think it's better and faster).

2026-03-05 20:58:06 UTC

- reply

it's beautiful

2026-03-05 20:28:53 UTC

> It’s been six years since GPT-3, and there has been no “move 37” moment for writing (as there was for AlphaGo’s creative play of Go). Not even close.
https://www.theintrinsicperspective.com/p/bits-in-bits-out

2026-03-05 18:36:45 UTC

This article says a lot of what I wanted to say about LLMs, but couldn't find the words: https://acko.net/blog/the-l-in-llm-stands-for-lying/
I don't agree with his conclusion that leaning in to intellectual property rights and source citation is the solution, though that's an interesting though. But there are some great sections, particularly in the first half. Here are some highlights:
> LLMs do something very specific: they allow individuals to make forgeries of their own potential output, or that of someone else, faster than they could make it themselves.
> Experienced veterans who turn to AI are said to supposedly fare better, producing 10x or even 100x the lines of code from before. When I hear this, I wonder what sort of senior software engineer still doesn't understand that every line of code they run and depend on is a liability.
>
> One of the most remarkable things I've heard someone say was that AI coding is a great application of the technology because everything an agent needs to know is explained in the codebase. This is catastrophically wrong and absurd, because if it were true, there would be no actual coding work to do.
>
> It's also a huge tell. The salient difference here is whether an engineer has mostly spent their career solving problems created by other software, or solving problems people already had before there was any software at all. Only the latter will teach you to think about the constraints a problem actually has, and the needs of the users who solve it, which are always far messier than a novice would think.

2026-03-05 16:25:19 UTC

- reply

Sounds great! Another feature request is to add api request history somewhere. On opencode I spent $20 very quickly and it was really helpful for figuring out exactly which API calls caused it. Including number of tokens in/out, dollar amount of upstream, and payment amount would be awesome. Also, I've noticed that the current tally of dollar amount doesn't show in opencode like it does for other providers.

2026-03-05 16:21:41 UTC

- reply

Nice, sounds like a great way to simplify. I like routstr a lot, but here are a few other things you could to do improve the UX:
- Don't process payments synchronously in chat
- Auto-refresh balances in the background?
- Maybe eliminate the balance on a key? Although having a budget on a key is nice.

2026-03-04 22:10:41 UTC

- reply

delusional

2026-03-04 22:08:44 UTC

Just added this to my opencode build prompt:
> You are in a docker sandbox, which means timestamps on files are often incorrect. To get around this, always touch a file before editing it.
🙄

2026-03-04 20:53:44 UTC

- reply

Yeah, pretty nice. Although it auto linked my email account, could there be an attack vector based on non-unique names?

2026-03-04 19:52:18 UTC

- reply

Got it working, thanks!

2026-03-04 16:54:06 UTC

@nprofile…uy5c https://git.mleku.dev/mleku/gitea-nostr-auth is giving me an SSL error

2026-03-04 16:50:56 UTC

- reply

Coracle isn't really the place to showcase dvms, custom feeds notwithstanding. Maybe something like noogle.lol would be better

2026-03-04 00:55:15 UTC

agents just introduce a whole new category of yak shaving

2026-03-04 00:55:03 UTC

Me: follows the directions for setting up NanoClaw on a fresh VPS
NanoClaw:
```
lsof /var/lib/dpkg/lock-frontend 2>&1
kill -9 19631 19750 2>&1; sleep 1; rm -f /var/lib/dpkg/lock-frontend /var/lib/dpkg/lock /var/cache/apt/archives/lock 2>/dev/null; dpkg --configure -a 2>&1
```

2026-03-04 00:52:19 UTC

- reply

How is sprites different from just setting up a VPS? How's the latency?

2026-03-04 00:51:13 UTC

- reply

I have been thinking about writing a client that would timestamp messages as sent on the current date in 2021 or something. could be fun, partitioning content by epoch. It would also screw up any idea of the authenticity of old posts

2026-03-04 00:46:34 UTC

- reply

You clearly never wrote php in 2012

2026-03-03 23:47:58 UTC

Vibe coding is the death of abstraction. Why use the visitor pattern or transducers when the LLM will just scatter `if` statements everywhere anyway

2026-03-03 19:44:57 UTC

- reply

That 20% matters. I try to use kimi k2.5 as much as possible, but it's just not as good as sonnet

2026-03-03 19:44:11 UTC

- reply

stuff != wealth

2026-03-03 18:42:50 UTC

- reply

I don't know, I don't really use this stuff. I can just smell the spaghetti from around the block

2026-03-03 18:42:04 UTC

Agents are getting more expensive:
https://newsletter.danielpaleka.com/p/you-are-going-to-get-priced-out-of

2026-03-03 18:21:51 UTC

- reply

No one should use openclaw, use nanoclaw or ironclaw or something else if you must run an agent

2026-03-03 17:42:52 UTC

- reply

#nevent1q…sa30

2026-03-03 17:42:37 UTC

```
docker sandbox save opencode-flotilla my-opencode:v1.0.1
Snapshotting image in sandbox ...
Reading image from sandbox ...
Save complete. To use the image:
docker sandbox create --load-local-template -t my-opencode:v1.0.1 [...]
```
```
docker sandbox create --load-local-template -t my-opencode:v1.0.1
unknown flag: --load-local-template
Usage: docker sandbox create [OPTIONS] AGENT WORKSPACE
Run 'docker sandbox create --help' for more information
```
great, thank you docker
#nevent1q…jl3k

2026-03-03 17:41:02 UTC

docker sandbox is clearly vibecoded. Half the flags don't work, and half the commands advertise flags that don't exist.

2026-03-03 16:38:31 UTC

- reply

what is UTM?

2026-03-03 02:51:14 UTC

Spent the day fiddling with agent isolation. At first I went down the rabbit hole of setting up a dev environment on an old macbook and acc,ssing it over wireguard, but the latency was annoying. Then I tried matchlock, which was promising but had weird build and control character issues. Finally, I went with docker sandbox, shich is good enough, although I had yo use a very dumb hack to get my config into the container. This is a massive product opportunity.

2026-02-28 01:41:51 UTC

- reply

They're not autonomous, I write the prompts

2026-02-28 01:07:56 UTC

- reply

Just coding

2026-02-27 22:12:32 UTC

- reply

nooo no no haha just opencode

2026-02-27 22:07:53 UTC

- reply

Setting up nix on an old 2012 macbook I have lying around. We'll see if it survives 😂

2026-02-27 21:30:55 UTC

- reply

Yeah, I was thinking about that. I intend to give it all its own keys.

2026-02-27 21:30:46 UTC

- reply

Do you ssh in and have all your code on the box, or send context from your machine to it somehow?

2026-02-27 21:08:50 UTC

- reply

I bought some old laptops from a school on ebay, 5 for $100. One of them doesn't work, but it was a great deal.

2026-02-27 21:08:21 UTC

Now that I'm using agents more extensively, I'm thinking about moving my development environment to a VPS to make sure the agents don't send any important data (like my ssh keys) to my provider. Am I being paranoid, or has anyone else done this?

2026-02-27 18:59:04 UTC

- reply

Very good articulation I hadn't hear before. We're already seeing a lot of those fragility points with the increasing ability to impersonate/scam people in ways that we have zero defenses against. Many more will come I'm sure.