Elkim on nostr

Recent Notes

GM https://blossom.primal.net/5d0b97aafe7f5b8763acdb788d96494dc2ca565d0f28e2767b278fd70ac1f0e8.jpg Tell me what book you are reading. I will zap from 0-100 depending how much I like it. #bookstr #...

Elkim @elkim 1788704316

❤️1🤘1

Elkim

@elkim 1788697533

Building complex open-source but niche apps with AI is fun from a start but after few months, it can get boring, little depressing and lonely.

Yet from time to time you finish super cool feature which brings back the joy from solving problems.

The hardest part is to get anyone to try the thing you are working on because everyone else is busy working on their projects. Attention really got pretty scarce with agentic AI and very capable models.

I wonder it it wears off and how the world will look in a year. Wild, exciting yet scary times ahead.

getbased.health @getbasedhealth · 1w

Huge update to getbased just landed. Use your favorite CLI agent as an AI provider. That means you can use your existing subscription with Codex, Opencode (including free models), Grok, Claude (self-hosteding only for now), Hermes, and OpenClaw. getbased now has its own tools Show more

❤️2

Elkim

@elkim 1787911990

Someone asked for this earlier and now here it is. Any nutrition-tracking AI lunatics here up to give it a try? :)

getbased.health @getbasedhealth · 2w

v1.18.0 — Meals & Nutrition arrives Track your meals, macros, nutrition and hydration with a vision model of your choice. Simply add a picture and run analysis. You can also use product labels or add everything manually. The precision of the analysis depends on the picture Show more

❤️1

Elkim

@elkim 1787809633

GLM 5.3 Flash available from, all providers. It's a new open-weight vision model scoring 91% on @ArtificialAnlys Scientific Reasoning.

Go try it, looks great!

Btw, new major feature dropping soon, stay tuned!

1❤️2

PayPerQ · 5w

Yea it's a worry!

Elkim @elkim 1786598857

I think it depends on a practice, I spend maybe 50 % of time building my project with refactoring and improving the codebase quality. I also try to target 90 % functional test coverage.

PayPerQ · 5w

PPQ's codebase increased 330% (48k-->159k lines of code) in the last 12 months. The rate of growth is both astonishing and alarming at the same time. I don't know the exact significance of this but i...

Elkim @elkim 1786543445

It doesn't matter in the age of agentic coding. Fat code ftw 😅

waxwing · 5w

It feels that way. But I suspect it'll come back to bite people later. Maybe some O(N!) scaling in the worst case.

Elkim

@elkim 1786116524

And bunch of recent updates incl. Venice E2EE verification to GPU, improved Cross Device Sync and a lot of other things.

getbased.health @getbasedhealth · 5w

A completely redesigned health context cards just landed - Health context has been redesigned from the ground up. The cards are cleaner, easier to scan and edit, work better on mobile, and explain why each area matters. - The questions are more useful and more complete. Medical Show more

Elkim

@elkim 1785907385

the fuck?

. @. · 6w

NVK knew in 2021. He told Matt Odell. https://npub1ak68qfcjj7k95c0jwleu69x72nr8adwv6g80pkwl9xlps6zmkqzqrxy8fx.blossom.band/200f65747d4b50e01a6e912a628e2c1da49fc4a45c5bfdccc2cb4fef783e8776.mp4

weev · 6w

Well, I gave the tribe the benefit of the doubt and simply called NVK a fat Mexican. But it turns out that literally 100% of all crypto losses are a result of this one ethnoreligious minority. Mark Ka...

Elkim @elkim 1785839799

FWIW Rudolf Novák is typical czech name, bit ancient these days since parents give kids woke Disney names but I think he might have roots here.

weev · 6w

Rodolfo definitely looks like a taco-guzzling spic, I’ve seen guys like him in the parking lot at Home Depot, he must be half spic and half Czech Jew

Elkim

@elkim 1785493402

Speech-to-Text and Text-to-Speech is out!

You can use:

1) Use local models Whisper for STT and Kokoro for TTS
2) Local server like Ollama or LM Studio
3) 3rd party provides xAI Voice or ElevenLabs

🎙️ http://getbased.health 🦻

Elkim

@elkim 1785318516

Totally missed that @getbased.health repo has over 100 starts. Kinda nice with almost zero marketing efforts.

I've been working on codebase improvements last few weeks and big changes to onboarding coming soon.

If you take care about your health, go give it a try, it's a powerful addition to your toolbox!

👉 http://getbased.health

1❤️2

Elkim

@elkim 1784556508

Built a little something for the local AI users in @getbased.health so you can choose reliable model enough for the use case.

Compare two or more models against a reference result or each other.

Import of 65-marker blood work report tested with native Qwen3-6-27b ) vs. BottleCapAI’s fine-tuned Thinkingcap model.

Thinkingcap was 36% faster with the same results.

You can also compare cloud providers and see how much you are paying for what speeds. Some dare to charge much more and are much slower as well 😎