Sabitlenmiş Tweet
Welcome Sury 🚀
Retweet is appreciated 🙏
dev.to/dzakh/welcome-…
English
Dmitry Zakharov
671 posts
Dmitry Zakharov
@dzakh_dev
ReScript enthusiast at @envio_indexer. Create tools from the future 🧬
Georgia, Batumi Katılım Mart 2022
50 Takip Edilen265 Takipçiler
Dmitry Zakharov retweetledi
Devs be cooking at the team offsite 🧑🍳
Once a month the team at @envio_indexer runs an internal hackathon.
Ngl… some pretty amazing things coming soon.
English
Met giraffes yesterday. Meeting with @envio_indexer team tomorrow. Exciting.
English
Dmitry Zakharov retweetledi
Now that adding code is easy, removing code is the new superpower.
English
Valibot shines in invalid data parsing! ⏱️ At 29 μs, it's over 3x faster than Zod (93 μs) and ArkType (97 μs) in our benchmarks. 😎
Fabian Hiller@FabianHiller
🚨 Missed our big announcement yesterday? We launched schemabenchmarks.dev! 🎉 An open-source project comparing top JS schema libraries like @zodtypes, @valibot, @arktypeio, Typia, TypeBox, @EffectTS_ and more. 🧩 Dive into metrics on download time, schema creation, and parsing speed. 🔥 Check it out now and see who's leading the pack: schemabenchmarks.dev/blog/welcome
Slovenščina
@poptyedev @apoorveth @dexbruce_eth @envio_indexer Especially since Solana RPC is a pain, and the shared post mentions a super expensive proprietary Helius endpoint.
English
@poptyedev @apoorveth @dexbruce_eth @envio_indexer Curious what's missing for you in HyperSync compared to Solana? In my opinion, the only trully missing thing is Solana support 😁
English
EVM RPCs lack behind Solana tbh
it's so easy to get the token holdings there with a simple call, but on evm you have to rely on external APIs
mert@mert
fun fact: I used no indexing for this Helius' new getTransactionsForAddress is fast enough that you can scan the entire history for an address at runtime!! since it has timerange filtering, you can split the history into N time horizons and fetch all of them in parallel
English
Dmitry Zakharov retweetledi
@0xRyanHolanda @envio_indexer There's some work done in that direction, but not really planned. A GitHub issue is the best to increase priority 😁
English
@envio_indexer Any plans for only allowing deployments when CI checks pass? 👀
English
👾 GitHub CI for tests in envio init
HyperIndex V3 projects created with envio init include a pre configured GitHub Actions workflow for running tests.
This provides a ready to use CI setup from day one, making it easier to run tests consistently as projects evolve and contributors are added.
English
In the next version of Sury - if I make it work 😅
45 failing tests by the way. Gradually progressing without AI. Unfortunately, it doesn't help much in my case.
The code in the screenshot is a successor to S.unnest and is already available for use. Made it ClickHouse style 😁
English
Dmitry Zakharov retweetledi
Miriad is hiring a @rescriptlang developer with strong UI/UX focus.
Also good opportunity to work on open source as well 🚀
👉forum.rescript-lang.org/t/miriad-is-hi…
English
@xanf_ua Очень жду когда появятся языки, ориентированные на ИИ. Экономные по токенам и без большого количества вариантов написания одного и того же
Русский
Чем дальше в AI, тем важнее читабельность кода.
Сегодня AI споткнулась в своих рассуждениях о вот этот объект конфигурации тест раннера:
{
...
stubs: { Popup }
...
}
LLM-ка решила что это стабает компонент Popup. Но нет, синтаксис объекта stubs (вот тот неудачный, это vue/test-utils если что) - { что-стабаем: true | [объект реализации] }
в итоге посколку это разворачивается в { Popup: Popup } - это как раз означаем "для стаба компонента Popup используес реализацию объекта Popup" - что означает "НЕ СТАБАЕМ ЭТОТ КОМПОНЕНТ"
Одна из сложных вещей для меня - бороться с замыливанием глаза, я-то привык уже к подобным финтам
Русский
English
Honestly wish I could pick "@envio_indexer engineers" as an agent in @cursor_ai . Unreal problem solvers. Zero hallucinations, all solutions.
English
Back then, I put maintenance difficulty at the bottom, but nowadays I think if LLM can't write your code, it's a bus factor alert.
So I decided to take a step back and write Sury architecture in a formal language first, making it as clear as possible.
English
Still failed, but this was the most productive way to use an LLM in a complex codebase.
I remember how, a few years ago, I highlighted: community, DX, performance, bundle-size, and maintenance difficulty as pillars of a good OSS library
...
Dmitry Zakharov@dzakh_dev
I'm going to try again, but use it more as a rubber duck and a support tool for debugging. I think it should be more productive. Also curious to see whether Wallaby MCP will be helpful in the end.
English
It was a power week. Claude shipped more commits than I did, and I like it 🤩
A goal for next week: Start using voice for prompting more. I'm not used to it at all, but it sounds promising for my wrist health.
Ah wait. I'll be writing an architecture spec instead, without AI...
English
My eyes started crying in the middle 🥲
Gergely Orosz@GergelyOrosz
Spent 3 minutes watching it; I could understand it even at the fastest speed (900 words/minute) Later could not remember a single thing from the 3 mins fwiw the most important things I learn from books/writing: I almost always not just read, but re-read YMMV of course
English
Shipped a new testing framework to @envio_indexer - in my opinion, the BIGGEST feature for the whole of 2026.
Agent-driven development is growing like crazy, and the ability to programmatically verify LLM changes is a must-have for any tool nowadays. And we are the first here!
English
The first attempt failed. I even used @wallabyjs MCP to provide LLM with more context from test execution. This was cool, but not enough critical thinking to suggest a proper fix instead of a lame patch.
Still bullish on AI when I know what to do, though 😀
Dmitry Zakharov@dzakh_dev
Time to try Opus for Sury. I really hope it'll work, because the logic got too complicated for my human brain 😅 Good that I have @rescriptlang and more than 1k tests to guide it through.
English