Andrey Yasko

Founder of @UNACMS and @YaskoStudio. Software designer. Interested in why people do what they do.

[Andrey Yasko @andyasko]

@swmansion Awesome! Finally a solid RN editor with Web supprot 👍👍👍

[Andrey Yasko @andyasko]

Looking good! - unacms/neo scored 100/100 on React Review https://t.co/VqhgZqT2oe

[Andrey Yasko @andyasko]

@wcandillon It’s the morus-glass example, though just tested and it works fine at 60 fps when playground is reloaded. May well be something on the landing page that overloads the thread and then examples lag. After reload it’s smooth as butter.

[Andrey Yasko @andyasko]

@Baconbrix @expo Oh, I didn’t even know you could do this in the simulator 🤦‍♂️. Thank you!

[Andrey Yasko @andyasko]

This is bigger than AI

[Cheng Lou @_chenglou]

1. Occlusion (virtualization) of hundreds of thousands of text boxes, each with differing height, without DOM measurement, therefore simplifying the visibility check to a single linear cache-less traversal of heights, scrolling & resizing at 120fps https://t.co/oAUwyHeTCm

[Andrey Yasko @andyasko]

@nikitabier a few suggestion:

1. Pure Mode

- Only posts and replies from verified and followed accounts.
- No “generated by ai” posts.
- No API posts.
- No posts/comments from accounts that ever made more than 100 posts/replies a day.
- No posts/comments from accounts that ever promoted any currencies or tokens.
- Quarantine new accounts in “Messy” mode until their posts/comments gathered 50 likes.
- Before adding accounts to “Pure” fanouts, run signal/noise validation review - it’s an expensive calc so you only do it once and voice them back if it fails. Get number of likes relative to posts, get a threshold Q where it’s more likely than not to be real. Also run replies through Grok to assess percentage of added value as opposed to rephrasing the OP (standard slop).

2. Stakes

- User can stake $10+ for each post they make. Use that as standard Boost ad budget, but credit back if post gets 10+ likes and is within your views/likes QQ (quality quotient) threshold.
- Raise min stake and price per view every time account fails to reach QQ.
- Make stakes compulsory for monetised accounts and min N percent of their estimated revenue per post.
- Reduce CPC/CPM for stake boost for accounts with higher QQ.

All that means more work for your Scala/Navi/filters/mixers, but you can recoup it:

- allow fee members to use Pure mode in read-only. Can like, bookmark and repost, can’t post, reply.
- don’t count likes from free members towards QQ.
- push ads/promoted posts to Pure mode if general QQ is reach in Messy.

This will cover more users to paid subs and will create more the “exclusive” curated X experience.

[Andrey Yasko @andyasko]

@shadcn Social good tech startup venture fund. Vest ideas that make positive impact even if not financially profitable - affordable housing construction tech, accessible lending, p2p gig platforms - anything that solves economic inefficiencies without cost of monopoly platforms.

[Andrey Yasko @andyasko]

@pmitu AI has all the knowledge, but no comprehension at all. Next frontier is in wiring up AI inference with human comprehension at scale.

[Andrey Yasko @andyasko]

@TukiFromKL Wait until AI adds a bunch of new functions following a hallucinated pattern that has nothing to do with the actual product logic.
Then it keeps iterating on its own slop until even code review passes because everything looks plausible.

It “works”, I guess.

[Andrey Yasko @andyasko]

@nikitabier Admit it. You must have prompt-injected all those clawds to spin up the counter. 😉

[Andrey Yasko @andyasko]

@Austen True and seems like should have always been like that.

Real concern is that all those bots can now reply via browser at scale. Looking forward to seeing that addressed.

[Andrey Yasko @andyasko]

@nikitabier It’s been insane. Especially those comments like … “if confirmed this represents significant escalation…” 🤮🤮🤮

[Andrey Yasko @andyasko]

@dagorenouf Correct. Also annoying as hell. They outsource scraping to 3rd-party bot farms, which then steal data but also render all analytics useless and add massive load to everyone’s servers.

[Andrey Yasko @andyasko]

Frontier LLMs set the culture for this themselves. Scrape the hell out of everything indiscriminately, copyright or not. If it’s accessible it’s fair play, right? That’s how Claude and CharGPT came to be. Now you expect someone to respect copyright of derivative data generated by your models?

Wait another year and most of the public published data will be LLM-generated anyhow.

[Andrey Yasko @andyasko]

@TaraBull

[Andrey Yasko @andyasko]

In Orthodox Christian sense that is the most important step towards salvation. True faith is predicated on consent, which requires informed choice. It can’t be predestined, coerces or even divinely cast upon someone. For an analytical mind it may well need to start with agreeing with teachings. For someone else it may require emotional connection. But the point is that Grace is given but not compelled - it’s there for you, but you have to choose it.

You start with intellectual agreement, and if choose so, add trust. That forms synergy between human free will and God’s grace, aka love in action, aka Jesus. The teachings, the embodiment and the tradition help to get to that point - they are there to inform, and if you agree and if you then choose to trust, then Salvation begins.

[Andrey Yasko @andyasko]

@mreiffy No real moat for tools. Social graph, proprietary data and realtime data will be the new defensible moat.

We build networks with agentic tools. Only raw data from human users will remain valuable.

[Andrey Yasko @andyasko]

@businessbarista Agreed. We pay for Linear not because we can’t create our own clone. In fact we could just download one. We pay because it’s reliable - gets updated, works consistently, data stays where it’s meant to be. Also for being sure that in a few years it’ll keep working.