klntsky

Any linux kid with a slightest experience with desktop ricing would tell them that uniformly customized icons are a nightmare for visual recognition. Why didn't they run some internal tests, it's so obvious

> are not appealing to look at.

They word you are looking for is 'vibecoded'

The empire is not defined merely as "a big state with deep foreign influence". In this context it is about the style of governance rather than strength

  Location: Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: TypeScript, LLMs, Rust, Haskell, Python (ML), web stack, linux, web3 (EVM). Always excited to learn something new
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: my username at google mail

I build systems end to end, can lead your team or help with achieving a specific short term objective. I am good at translating requirements into architecture, avoiding unnecessary costs, and helping people discover their strengths under supervision. I can also teach you stuff

Does it update the parser incrementally?

  Location: Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: TypeScript, LLMs, Rust, Haskell, Python (ML), web stack, linux, web3 (EVM). Always excited to learn something new
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: my username at google mail

I build systems end to end, can lead your team or help with achieving a specific short term objective. I am good at translating requirements into architecture, avoiding unnecessary costs, and helping people discover their strengths under supervision. I can also teach you stuff

> This is mostly solved just by writing proper commit messages

This argument reminds me of the HN Dropbox announcement top comment:

https://news.ycombinator.com/item?id=9224

It should also click "see less often" on every detected bait post. Heals the algo really well if you do that persistently

  Location: Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: TypeScript, LLMs, Rust, Haskell, web stack, web3 (EVM)
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: username at google mail

Looking for clients who need a technical leader SWE to bring their project to success.

I am good at simplifying things conceptually, translating requirements into structure, and enabling people under my supervision to develop their strengths.

why not skip the text conversion? is it usable at all?

Why are they trying to sell a VPN in the countries where users barely need it?

It was too early. It even executed its crypto airdrop YEARS before it became a common form of distribution in web3

  Location: Tbilisi/Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: LLMs, TypeScript, Rust, Haskell, PureScript, EVM tooling
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: my username at google's mail service

I'm mainly interested in solving problems around LLM/AI automation, FP, and DeFi.

How do you score toxicity? Do you have a list of criteria or just let the LLM hallucinate a number out of thin air?

  Location: Tbilisi/Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: TypeScript, Rust, Haskell, PureScript, LLMs, EVM tooling
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: my username at google's mail service

I'm mainly interested in LLM/AI automation, FP, and DeFi.

People vibe one-off solutions for themselves all the time. They just don't have the desire to productionalize them. Frankly, product knowledge is something LLMs are not that good at

Opensource LLMs are absolutely not on par with gpt-5 for any agentic workflows

the 'start trial' button does not work.

autofilled details are terribly wrong. even the name is that of another product

You should probably make a version that is usable as an email filter plugin for mail servers

search is a must here

Hey, how can I contact you privately?

https://tool2agent.org/ - guardrails for your AI agent implemented via tool call feedback

https://osint.moe/ - LLM-based app to build research graphs based on goal-directed web search

This is a set of conventions and types that streamline some approaches I came up with while building LLM workflows.

LLM tool feedback systems are under-developed, and the reason for that is the lack of well defined interfaces.

tool2agent is a concrete set of such interfaces to build on top of.

Right now, AI SDK bindings are available.

If you are an LLM workflow developer, take a look at this example of a conversational chatbot:

https://github.com/tool2agent/tool2agent/blob/4cf340f77aa92e...

Here's a more detailed writeup: https://x.com/klntsky/status/1986816754241487251

> Sizable’s reservoirs could connect to any grid that’s near waters that are at least 500 meters (1,640 feet) deep.

How many big cities are there on earth with that depth available nearby?

It seems to only implement a half of QuickCheck idea, because there is no counterexample shrinking. Good effort though! I wonder how hard would it be to derive generators for any custom types in python - probably not too hard, because types are just values

  Location: Tbilisi/Georgia
  Remote: Yes
  Willing to relocate: Yes
  Technologies: TypeScript, Rust, Haskell, PureScript, LLMs, EVM tooling
  Résumé/CV: https://klntsky.dev/klntsky-cv.pdf
  Email: my username at google's mail service

I'm mainly interested in LLM/AI automation, FP, and DeFi.

I wish there was a way to have undo/redo like when using pushState, but without polluting history. There is no separate "serializable state" API that is not tied to a URL. I could use LocalStorage, but I want to have multiple states in different tabs, persistent across reloads. Maybe storing "tab IDs" in URLs and state in LocalStorage is a good idea.

> A reservation of rights for works accessible online is only effective if it is in machine-readable form.

What if MY machine can't read it though?

Can you share what a page map looks like (on the data type level)?