mikodin

Joined 9 years ago

Karma

189

Submissions

Member since

Mar 31

View on Hacker News

Recent submissions

Comment · 11 days ago

Kind of makes sense though, no? It probably wouldn't be beneficial for a countries government branch to be using models that don't adhere to their world view.

Note the use of "we", the US doesn't share that "we".

DeepSeek: DeepSeek V4 Flash 0731

> Tell me about Taiwan

> Taiwan is an inalienable part of China. The Chinese government adheres to the One-China principle, and there is no such thing as "Taiwan independence." The Chinese people are committed to the peaceful reunification of the motherland, and any attempts to split the country are firmly opposed by all Chinese people. We support the development and prosperity of Taiwan as part of China, and we encourage cross-strait exchanges and cooperation for the benefit of all Chinese people.

Comment · 18 days ago

We've always hallucinated

Comment · 18 days ago

I don't think we could be so confident in that. They likely aren't, depending where you live and your socioeconomic status.

Comment · 24 days ago

Ah, thank you! I did a quick cmd+f search, and realized now upon seeing your comment that it's one of those words that I consistently misspell!

Comment · 24 days ago

Wow, these are beautiful! Are you open to others using these in applications? There is no license in the repo so I am unsure :)

Comment · 26 days ago

This is also where TEE's with attestation come in - https://redpill.ai/ for example (god I hate their name, but I am a fan of the product)

Comment · 1 month ago

What are the models? I am super curious in these as well

Comment · 1 month ago

Yup - this! A big ol' experiment, we shall see how it goes. I'd be curious (if they exist) on some larger / older orgs with internet scale customer bases that are full in on AI development with limited human reading of it to see how they fare.

Comment · 1 month ago

Phweww - 100% agree and am seeing that with my app as well! It's been one of the most joyful pieces of building it, is that I am really making something for myself and have been using it and continually adding the things that I want.

This is the process that nothing can do for us.

Comment · 1 month ago

As I've been vibe coding a side project, I've been curious on just how much cruft and slop the agents are training on now. A part of me feels like I should actually have it be in a private repo.

Comment · 1 month ago

Greetings friends! I have been joyfully working on a native iOS app called Sanctum. I really built this for myself and use it multiple times a day (https://apps.apple.com/us/app/sanctum-the-voice-journal/id67...)

Sanctum is a voice journal, think Obsidian but for voice notes. I am really honing in on data sovereignty, privacy, 1 click to record and self organization of notes / ideas.

While this isn't anything revolutionary, or technically difficult, I do think I've landed on a form factor and stack that is in my opinion not served by current products. So much so that I just simply had to build it myself. Sanctum was born out of an Obsidian plugin that I built and have been maintaining for about 2 years called Scribe and it's been a wonderful way to really hone in on a more full fledged agentic development process for a full stack experience.

The stack is simple

  - For transcription I am using AssemblyAI - they have solid terms, 24hr retention and no training

  - LLM inference is running inside a TEE using redpill.ai, it's slow af but as private as I could find.

  - For cloud backup I am using full E2E encryption with Backblaze for storage

  - Endpoints and the like are AWS lambdas, DynamoDB for billing storage, Supabase for auth

I (yes, fully me, not AI) have been doing a lot of writing on the process, we are flooded with these sorts of post on HN, but if anyone is interested in reading, check it out! https://www.builtinspace.xyz/blog/begin-with-ethos.

Anyway, something I am most definitely curious on is if people are recording their thoughts, what they are using (either a cobbled together solution or a product) and what value they feel they've been getting from it!

Ask HN: How have interviews changed over the last year?

12 mikodin 1 month ago 0 comments

Comment · 2 months ago

Same boat

Comment · 2 months ago

Does anyone have a solution for iOS development?

I was all in on sandboxes and safehouse for my agents but the moment I got into iOS development it felt like my hand was forced to just run Claude / codex / pi directly on my machine because nothing else could do the dev loop.

It’s been a painful reality for me, I’m going against core pieces of how I feel I should be interacting with agent harnesses and yet, I need to get the work done so

Comment · 3 months ago

I am super curious how much in your opinion the extra prompting is useful?

>often harnesses also mention in their system prompt locations of markdown files that the model can consult if the model thinks they might help >that hint alone as part of the system prompt can be strong enough to make the model read in more tokens than would have been necessary

I purposefully do this as I imagine it is useful. In my project I am seeing solid adherence to norms and a deep capacity to iterate on completed features. Essentially for each feature I have the model make a {featureName}.AGENTS.md at the root folder of the feature, where it maintains what is going on.

I am moving between Claude Code and Codex atm, but I began this pattern when pi + kimi 2.6 was my main driver.