Aedifico

Public Key

npub1h3949ndf5ty2zlgaqd65gad2w5kvcax6jpqm8knms39su3ldswysha728v

Profile Code

nprofile1qqstcj6jek569j9p05wsxa2ywk482txvwndfqsdnmfacgjcwglkc8zgppamhxue69uhkummnw3ezumt0d5q3vamnwvaz7tmjv4kxz7fwdehhxtnnda3kjctvdn5hrz

Publishing to

Author Public Key

npub1h3949ndf5ty2zlgaqd65gad2w5kvcax6jpqm8knms39su3ldswysha728v

Show more details

Last Notes

2026-07-17 15:08:40 UTC

Fun building a local first app with a dumb encrypted relay, and bricking the local database 🤣🤣🤣

2026-07-16 14:17:09 UTC

- reply

If you are talking about in-transit emcrypted data, sure. What I meant is, if each phone becomes a FIPS node, and they have internet, bluetooth, wifi to connect to other nodes, it has a bigger reach of devices, so possibly bigger damage is possible. But not inherent from the design.

2026-07-16 13:46:33 UTC

- reply

Would be nice and dangerous too 🧐 If its a fips node, viruses can also propagate better through the different transfer mechanisms. Like the zero day series. But I dont say it is not already possible with apps million people use though.

2026-07-16 12:43:21 UTC

- reply

I love the versatility of your attention Kat! 🫂

2026-07-16 12:41:47 UTC

Actually it is quite amazing that you can be an app dev/android dev/ios dev if you have the persistence to it, just in a few days, and you can learn so much.
Learning by doing is the new norm. Lightning fast.

2026-07-16 12:37:24 UTC

- reply

I would vote for FIPS first, and then sell servers and compute over FIPS.

2026-07-11 14:27:57 UTC

- reply

Whats your "proper remote setup"?

2026-07-10 05:40:04 UTC

- reply

Experiences so far?

2026-07-09 08:23:15 UTC

- reply

Interesting take! What would you attribute this to? What advantage that allows him to overtake others?

2026-07-08 10:21:07 UTC

#HealthOS in-app agent at the moment can generate a personalized mealplan for $0.09 with gemini 2.5 flash.
Would you pay 9cents for a mealplan tailored for your nutrition needs and recipes?

2026-07-04 18:48:35 UTC

- reply

I think it worth to learn a lesson from it.

2026-07-03 06:44:51 UTC

Music is amazing. It can calm you, relax you, pump you, hype you, smile you...
great good morning ☕

2026-07-01 09:16:50 UTC

- reply

or more like privacy by default.

2026-07-01 09:16:19 UTC

I think it is an unheard thing that when you share data with an app, you dont necessarily want to share it with the company.
E.g.: I want to record my health signals with a wearable, and I use its own app. Why does using the app necessarily mean I share with the company? Why couldnt I record the signals, see in the app, and when I need something that cannot be calculated inside the app, I can decide to share with the company?
This sounds privacy by design for me.

2026-06-30 20:22:45 UTC

- reply

Wild times.

2026-06-30 19:32:37 UTC

- reply

Be careful, because opensource models soon will become dangerous.

2026-06-30 14:56:20 UTC

I think in the future,
2 year old kids will just run AI agents better than their parents/grandparents,
like today kids with mobile games.
Somehow kids these days are born with the phone in their hands. Not a good direction if you ask me.

2026-06-30 14:47:43 UTC

- reply

Or typing words in a calculator with numbers 😂

2026-06-30 14:44:52 UTC

- reply

- following plan, eating 1 portion -> tick (no measurement)
- photo / voice -> if not following plan, or extra
- using average food/fruit/vegetable sizes -> I ate a small tomato
Do you use an API key in your app then? I guess for these, context is small, so API cost is not huge.

2026-06-30 14:39:25 UTC

fixed mindset and growth mindset people are quite apart

2026-06-30 08:27:44 UTC

- reply

I guess thats one part of it. Have not tried it yet.
But I also dont want to take a photo of everything I eat. Plus I want to share as less data as possible with proprietary models. So I am working on using local models in the app. Not sure how well they estimate.
Does your app do photo estimates? How reliable it is?

2026-06-30 05:33:05 UTC

I am working on a way to track my nutrition, without the hardships of tracking.
Usually tracking is tiresome, you have to measure things, if you eat out, its hard to measure.
Lets change that.
I will make #HealthOS the king of low-effort tracking.
Have some part in-the-works for that.

2026-06-29 10:05:02 UTC

- reply

Privacy, what is not shared, cannot be stolen.

2026-06-29 09:44:46 UTC

It is quite cool, my #HealthOS tracks my health signals (movement, sleep, HR, etc), my nutrition intake, my pantry.
So I can leverage each information where necessary:
Shopping: meal plan - pantry
Calorie need: Base Metabollic Rate + Movement Signals
Nutrient need: base info
And this is only the beginning.

2026-06-27 10:56:44 UTC

- reply

I think the sadest part is, that if you buy a wearable device, and your are mostly trapped with their own proprietary app.

2026-06-26 17:40:39 UTC

- reply

What is the difference compared to cavemen skill?

2026-06-24 05:33:29 UTC

#HealthOS nutrition tab is mostly complete
- Recipes/Foods database
- Recipe import from url
- Food import from USDA/OFF (+with barcode)
- Meal plan view
- Nutrition view
- Pantry to track food at home
- Pantry aware shopping list for mealplan
- Cooking plan to know when to cook
- Recipe cooking mode
- plus some small niceties
Lets go!
https://blossom.primal.net/21c6abd3afd8b7c6477bdeb66c0665340308a555d253c4eb04e9c0568bc93c25.jpg

2026-06-23 05:39:01 UTC

- reply

Nice! Have you thought about a "text only version" for this, and make this view have a bit more life with animation/gif, and a shareable gif version?

2026-06-23 05:36:27 UTC

- reply

Isnt great experience, when you dont need support?
bugfree code, no support are the basis of a great product. All achievable of course... 😂🤙

2026-06-23 05:17:52 UTC

@nprofile…nsnk is a nice idea. Do I understand it correctly that it needs publicly visible ips to route in a cross-network situation? So CGNAT is a failure point?
#asknostr

2026-06-23 05:03:49 UTC

- reply

A good analogy is when you delegate a task to another person.
- How do you do it?
- How do you check if he is done?

2026-06-22 10:35:13 UTC

- reply

Do you feel you changed in the last couple of month? That something clicked?

2026-06-22 05:43:42 UTC

- reply

clank-aholic 😂🍻

2026-06-21 07:24:17 UTC

- reply

Great plan 🤙
https://blossom.primal.net/153c09076dc42df93444fc8454dd798802235178cb3774654bb7a7ec84d0b84e.jpg

2026-06-20 13:59:02 UTC

- reply

print("debug")

2026-06-20 13:53:00 UTC

- reply

I can just move on.

2026-06-20 13:51:55 UTC

Life motto:
I can fix my broken things.

2026-06-20 09:34:13 UTC

- reply

I am tinkering with the idea of ai-app-sandbox.
You have an app frame with a chat interface for any model. The model can use local tools to update the app-sandbox with the features you need.

2026-06-20 08:31:27 UTC

I figured that building for the expectations and needs I have is completely different that the masses have. So my freetime buildings wont target the masses, only my work I want to sell.
Would be good to do the two at the same time, but it does not seem efficient at all.
Where I can "merge" the two, I will, but not targeting it anymore.

2026-06-20 08:25:26 UTC

Life is awesome, when you figure how can you just build things.
great good morning

2026-06-18 05:30:43 UTC

- reply

Best customer ever. Pays. Gives no work. Not complains.
You deserve a badge or something.

2026-06-17 05:09:23 UTC

- reply

You cannot discriminate other formats!!!!

2026-06-16 13:39:55 UTC

- reply

I think everything shall be quotable, zapable, commentable. Could turn out quote useful, you just need a nice rendering.
Zap a comment on a zap of a vote.
It sounds an everyday usecase.

2026-06-16 05:16:49 UTC

- reply

If you strictly going for achieving the what, without the how, more things might be possible. Also many things are too complex to worth it.

2026-06-15 05:50:54 UTC

- reply

Competition to the rescue

2026-06-15 05:33:20 UTC

When you use sovereign/privacy aware tools, in many cases, if you want to develop something for yourself, but want make money with it or make it mainstream, you will face the issue, that the tools you want to use yourself vs people could use are really different. That you have to completely make a "different app".
Right @nprofile…jejf?

2026-06-14 07:05:40 UTC

- reply

AI scan from photo?

2026-06-14 06:44:08 UTC

- reply

Actually linux + local/remote AI combo, will be huge

2026-06-14 06:42:20 UTC

- reply

How did you solve the logging of food issue?
I found logging food not that comfortable, and longterm not maintainable, so I made my food tracking done from journaling.
You just write a daily "journal" with words what you ate, can take photos of the products and their nutrition table you ate. Plus it can use the openfoodfacts database for items without a photo.
Works quite nice. As done by AI, it is quite flexible.

2026-06-13 08:10:03 UTC

Did anyone face this issue, that the DNS is leaking on Android with a correct upstream wireguard config?
#nevent1q…wglv

2026-06-13 06:29:04 UTC

- reply

What dev stack did you use?

2026-06-12 05:11:05 UTC

- reply

I guess expectation vs reality is the issue with trust too.

2026-06-11 08:26:56 UTC

- reply

I think it is too fragile, but lets goo!
My take on it is, I started building an ai setup tooling (simple folder with slash commands), that creates an ai team for your usecase with many ai harness. Then if it works, I could have a UI around it actually.

2026-06-11 05:40:14 UTC

- reply

Then use one for only work. If you force yourself to work, you might come up with ideas and fixes how to spend less time on it, than doing the fun. Or you will just skip work like me 😂

2026-06-11 05:35:40 UTC

- reply

Have 2 subscriptions, one for work, one for fun. If fun out, work with the work one.

2026-06-11 05:31:39 UTC

- reply

My finding is based on research and experience is, that not more autonomy but better system design for the agentic system, that makes the difference.

2026-06-11 05:29:57 UTC

- reply

I have a setup, which I am fine tuning, that rather solves stories, not apps at this level.
It can work end to end in a sense, that it has the whole loop of deving in it, with independent verification, review and fix.
But I am reviewing its plan before the implementation loop.
I did a simplification round yesterday. Now I have an orchestrator and 3 subagent (implementer, verifier, reviewer), and then the orchestrator just manages the loop.

2026-06-11 05:22:38 UTC

- reply

What do you want to achieve with it?

2026-06-10 19:31:46 UTC

- reply

Like Jameson original? Which is actually not the original, but flavored in a way to suite more taste buds, and could sell in America too?

2026-06-10 19:25:44 UTC

Does nostrvpn on android ignore the dns conf in wireguard upstream conf for you too?

2026-06-09 19:19:01 UTC

- reply

Yepp, yepp

2026-06-09 08:24:09 UTC

- reply

You do great!
great good morning, wish you a great day

2026-06-04 13:39:56 UTC

- reply

True! I guess if you go to a random coffee shop and demanding to pay with lightning you will have low success rate. Also from the point of view of a business, you can also sound like evangelizing. I have no clear solution yet how to effectively spread, in this regard, what you want. It also highly depends on the openness of the other side, plus how much prejudice they have.

2026-06-04 13:10:22 UTC

- reply

My strategy since a while is, I dont want to tell others about nothing, until they ask, and I am minding my own business, using what gets me what I want. The point I want to make is, I dont think its a good strategy to care about other people in a sense, that you need them. In short:
Build and use the things that works now, and does not need new people, and evolve from their.
This is a winning strategy from day one. I am oversimplifying of course, but I think, this is directionally useful.

2026-06-01 15:40:04 UTC

- reply

I meant an agentic team. I have 3+ teams, only one for coding.

2026-06-01 12:02:31 UTC

- reply

I am testing a similar idea in the next weeks.
So far I have an adversary reviewer, that works better than a simple reviewer, and complains the hell out of the code. 😂

2026-05-31 08:36:16 UTC

- reply

Yeah, comfort vs efficiency.
You will always need to be a hackster to be efficient, because to make comfort for everyone, you will have to sacrifice a lot.
Thats how claude code and such try to be good from the beginning by setting up the system prompt, so it works for literally everyone.
Comfort needs generarility. Generality can never be efficient.

2026-05-30 20:01:19 UTC

I am tinkering with the thought of hosting my own agent/LLM since a long time.
I did not have yet too much success. Maybe its time from the signs, to try again.
I think the future is bright to use frontier models for creating new stuff/stuff without personal data/stuff for your local agent teams. Local agent teams to use your data, automate your personal stuff.

2026-05-30 19:55:59 UTC

My improvement loop now with my agentic team is quite simple.
I built an agentic AI five pillars/best practices document.
When an agent or the team does something badly, I ask it to audit the current agentic system compared to the five pillars, what caused the issue, and tell me how to fix it with the exact new content and its location.
Works great so far.

2026-05-30 18:40:32 UTC

- reply

The not-so-little little things...

2026-05-27 19:50:18 UTC

https://blossom.primal.net/87ad1ab80ff4cca04acc487079883ca2464d165367213a72b0cc6f6c88f975a6.jpg

2026-05-27 19:46:29 UTC

https://blossom.primal.net/06caff4d0e3534f809842a431eaf18c39eecad5476b7352073e4cebb26dc13c0.jpg

2026-05-26 10:38:22 UTC

- reply

Changing assumes you have a "fixed form". 😂😅🤙🏞️☕
Its like saying to a river it changed, however at each moment when you see it, its different. So is it changing?
Or just refusing be the same?
Sorry, too much coffee. ☕

2026-05-26 10:02:52 UTC

I am developing a pocket AI team. Its a PWA so you can access the power of your AI team anywhere.
I am not doing fully active work on it, because its a fun project, but I have a business to start. So dont expect a fancy tool too fast.
great good morning

2026-05-26 06:47:48 UTC

- reply

Simplicity and repetition are great tools 👌

2026-05-26 06:29:49 UTC

- reply

retest

2026-05-26 06:26:30 UTC

- reply

The aiming was a bit off 😂

2026-05-25 09:58:33 UTC

Its crazy that...
Nah, I just wish you all a great day, its coffee time here! ☕🏞️🤙
#nevent1q…342t

2026-05-25 09:23:27 UTC

- reply

I think people like things because of the feelings it invokes in them.
That is why you watch different movies and series, to live the feeling you crave in that moment.

2026-05-25 08:17:50 UTC

- reply

https://blossom.primal.net/0cc6a5853c4e7c1ad0b0a20fce46f80cdd714aaddbcfe5fe5b18e48f121a0297.jpg

2026-05-23 18:35:49 UTC

- reply

If you want to reuse it, ask the AI to make a script to automate.

2026-05-23 18:35:39 UTC

- reply

It will be faster.
Not everything worth a tool or an app.

2026-05-23 18:34:47 UTC

Use AI as a tool to reach the goal, not to build a tool for a goal.

2026-05-22 06:12:03 UTC

- reply

Great good morning ☕🏞️

2026-05-21 11:47:07 UTC

- reply

True! The new sparrow seems to be shipped with a public frigate server: https://frigate.2140.dev/
#nevent1q…5nss

2026-05-21 06:10:33 UTC

Thats interesting, and kind of efficient to build system with AI where the same AI is part of the system.
So basically AI does the job, and then you do retros with the AI to do the job better.
It is a lesser complex idea, than building a tool with an AI, that has to produce an output. Because there you have 2 outputs: the tool and the tools output. So you have to improve your AI, your tool, and your results. 3 layers, instead of 2.
But if you cut out the tool layer, you kind of simulating the manually doing something instead of automation.
And actually this is how you automate properly. You define first the process, fine tune the process, then automate. This way, you will be faster reaching your result, and fighting less fights.

2026-05-21 06:03:41 UTC

I am aiming to improve my health with data.
So I have an AI team to research, build, measure, reason about data.
My current stance is, backed by data, that you shall pay attention to food variety but also ingredient amount.
So I have now a table that summarizes my nutrients goals and variety goals for my generated meal plans. Both paint a different picture.
You can eat in great variety, but still overeat, undereat, etc. Variety + the right amounts matter.

2026-05-21 05:51:07 UTC

- reply

Always, enjoy! Great good morning! ☕

2026-05-21 05:32:52 UTC

Who would be interested in a project, where you run AI teams in your browser?
No server, just purely in the browser, your data stays local.
AI team for
- personal development
- nutrition/meal planning
- sport
- business administration
Thinking about a webapp, what runs locally, can download teams locally, and then the data stored by the team stays local, stays in your browser, can be downloaded.

2026-05-21 05:26:34 UTC

- reply

I am not sure if it is wise to give people what they did not ask for, do not understand, and might not even be able to use it (get it out).

2026-05-21 05:23:08 UTC

AI agents == flexible workflows
They dont have 100% deterministic output, but also do not need 100% deterministic input.

2026-05-21 05:22:09 UTC

I advanced yesterday my coach team. Its quite awesome.
The "meal-planning squad" can now:
- plan meals from my recipes
- add random filling ideas from non existing recipes
- automatically registers new ingredients
- compares the weekly plan against variety goals and nutrient goals
It is quite simple though, bunch of md files, I sync across my devices, and a python script to run some deterministic work.
No app, no database, no hard dependencies yet.
Its so flexible (because of AI), that it can adapt easily to new ideas.
I am planning to go from AI to semi-AI, and make most of the things automated, so flexibility stays, but uses less tokens, and be more deterministic.

2026-05-21 05:13:32 UTC

I made yesterday a small visualization of my work llm credit spending.
I wish I have done it earlier.
great good morning

2026-05-20 14:52:27 UTC

- reply

It should definitely be an option.

2026-05-20 09:49:09 UTC

Imagine you download a social media app and upon signing up it opened a bank account for you. 😂😂

2026-05-20 05:37:59 UTC

I am wanting to build a meal planner app, that plans your meal around the goal of healthy nutrient ranges.
My last attempts all were facing a lot of issues in a sense, that it would have to bridge many hard gaps, so I always started easier, and not meeting the original goal. I learned and improved with every trial.
The last trial, when I made an AI coach team to be the app was the most successful.
The whole app-experience is mainly markdown files plus 1-2 small python scripts, but achieved 100x more in a few hours as ever before.
The learning:
You can use AI to build an App. You 10x your output. If you make the AI do what you need, you shrink the improvement cycle, and 100x the output.
The cool thing is, that you can prototype an idea by the AI doing it, then automate the steps.
Then you dont need to implement a single line of code until you know this step makes sense. As it should be.
great good morning

2026-05-20 05:27:16 UTC

- reply

Nice! I did some tracking in the past. It works nicely.
My longer standing idea I have now is:
The plan is the tracking.
+An AI coach team is the app.
You dont log, you plan, and the plan is what you ate. Less adaptable to free style eating (but AI could also handle), but I track nutrients as well.

2026-05-19 18:54:39 UTC

- reply

Rich for a second, poor for another.

2026-05-19 16:00:46 UTC

- reply

#nevent1q…0ree

2026-05-19 13:03:35 UTC

I like the virtual filesystem shakespear diy has.
We could make a "local" in the browser AI team with it.
You could talk to your team even through your fridge. (Not endorsing fridges though)