dan_voronov

I’ve been researching how LLMs generate code since summer 2023. My main public artifact is this repository: https://github.com/danvoronov/CodeWithLLM-Updates

It’s a living research collection with 400+ commits covering practical experiments, usage patterns, step-by-step guides, model comparisons, AI coding tool evaluations (Cursor, Aider, Cline, Windsurf, etc.).

A large part of my work is *exploratory*: navigating large and messy information spaces, testing things hands-on, comparing different approaches, and identifying the small amount of signal that is actually useful.

My strongest skill is broad technical *investigation*. I can map an entire problem space rather than focus on a single tool, model, or hypothesis: find the relevant projects, papers, tools, workflows, discussions, benchmarks, and practical examples; test and compare what matters; and organize the results into a clear picture of the field.

The goal is not to arrive with a predetermined answer or recommend one particular tool. I can cover the *surrounding landscape* as thoroughly as possible and provide the evidence, comparisons, and structure. Your team can then decide which conclusions matter and what to do with them.

I can deliver the results as a structured PDF or HTML research artifact, or discuss the findings in a call. Depending on the project, I can also combine both formats: prepare a written report with the relevant evidence and comparisons, then walk your team through it and answer questions live.

I’m interested in projects where someone needs a broad, hands-on investigation of AI-assisted software development: understanding what exists, how different approaches compare, where they work, where they fail, and what is worth paying attention to.

I don’t know the conventional job title for this niche. If your team needs someone who can cover a large and changing technical field, turn scattered information into a structured overview, and give you a solid basis for making decisions, I’d be interested in talking.

Who: Dan Voronov Location: Kyiv, Ukraine Remote: Yes Github: https://github.com/danvoronov Email: dan.voronov@pm.me

I suffered with Antigravity 1 and 2 for a long time (I’m not some weirdo, I just run a blog where I do deep dives into AI tools for programming=). The desktop version was basically unusable, so I used the console one. A couple of weeks ago I finally dropped it and switched to Pi with antigr login. With plugins you can configure it however you want. I just installed one of a deep-search plugins too: you need to add API credits from a search engine, but it can run multiple search rounds with a specified depth.

Compared $20 ChatGPT and AGY plan: my weekly limit on sol/terra used to disappear in one intensive day (or two normal ones). Now they’ve added the 5-hour limit on top of that, which is even more annoying. With Google I barely notice the limits at all — the only one is Opus 4.6 (not sure why it has one), and I only use it for double-checking sometimes.

The biggest problem is that the Pro model hasn’t been updated, and Flash just doesn’t produce solutions of the same quality as Sol. But it is matter of time...

I absolutely love the idea behind Shellcast.tv – "Stream your CLI coding" is a fantastic concept that the community really needs. However, I was very disappointed to find that the current implementation is unusable.

It's a shame because the potential is huge, but it feels like the project was released without any basic quality assurance. Here are the critical issues I encountered:

- Windows Installation: The installation command provided fails with an error right away. - Google Login: The Google authenticatio in my FF is broken. After granting permissions, the page just goes to a blank white screen and never redirects back to the application. - No Discoverability: I couldn't find any links or a directory of currently active streams, which defeats the purpose of the platform.

In its current state, the service is unfortunately not functional. I sincerely hope you take this feedback to heart, perhaps bring a QA engineer onto the team, and work on delivering a stable, working MVP. I'm rooting for you and would love to use Shellcast.tv once it's ready.

GOOGLE: The `photoslibrary.readonly`, `photoslibrary.sharing`, and `photoslibrary` scopes will be removed. API calls relying only on these scopes will return a `403 PERMISSION_DENIED` after March 31, 2025. Several other calls will be restricted to working only with photos and videos created by your app.

I couldn't even log in, which is why I really disliked this tool. The only thing I was able to do was change the color theme.

When I try to log in using a personal account, it tells me I need to generate something in the dashboard. I know that on my main account from maybe 15 years ago I had things like domains thats nowdays Google Workspace. But I'm trying to log in and authorize using my second (NOT MAIN), relatively new account. Despite this, it redirects me to the console (like it would for the first account), and there it says I'm logged in (!), but the application itself says I'm not.

I switched profiles within the console too, tried the same steps, and it still results in the same problem. Yes, I can see my new profile in the top right corner, not the old one. It sends me to a "Users may have to specify a GOOGLE_CLOUD_PROJECT if:" github [0] saying this ID is needed, and I honestly would even generate it, but I don't want to waste time with the program if you can't even get basic authorization working correctly.

0: https://github.com/google-gemini/gemini-cli/blob/main/docs/c...

Big plus for supporting Firefox!

There is a compatibility issue with the "Modern for Hacker News. A redesigned web interface for Hacker News." browser extension (I use Firefox). If it is active, I do not see the block with related news.

Can this be fixed?

Thank you very much, it looks very cool. Personally, I miss the *hotkeys* because it's not very convenient to constantly click "show" on the automatically collapsing settings block.

The idea is very good, it's just annoying that the current (i made ai startup) trend is to vibe generate something that is a wrapper for other AI models and already ask for $15 for it. Do you even realize how much work has been done in OpenAI to ask us for $20 a month? How many engineers and scientific articles?

From my point of view, for this not to be a scam: 1 - there should first be a video presentation of the program's operation with all functions, not a website with a bunch of promises and a PAY button 2 - You should definitely write your FIRST startup without any AI generation at all!

Hi there,

First off, I want to say that the core idea behind this – exploring locations through Street View and providing contextual information about what you're seeing – is absolutely fantastic! That's a really unique and engaging way to learn about places, and it has a lot of potential. Great concept!

Now, regarding the user interface... to be honest, while the concept is strong, the current look and feel of the interface feels quite cluttered and dated. :( It feels like an "Indian style," and I think what you might be picking up on is a design approach that tends to pack a lot of information into distinct boxes with varying styles and colors, sometimes leading to visual noise rather than a clean, organized presentation. It doesn't strictly adhere to modern design principles of simplicity and consistency.

Here's a breakdown of why it feels less modern and what could be improved:

Why the Current UI Feels Dated/Cluttered:

1. *Inconsistent Visual Style:* There are many different box styles, border radii, background colors, icons, and font treatments all over the screen (the purple text box, the varied colored boxes on the right, the Explore tags, the buttons). This lack of uniformity makes it visually jarring and hard to process quickly. 2. *Poor Visual Hierarchy:* With so many prominent boxes and varying colors, it's not immediately clear what the most important information is. Everything seems to scream for attention equally. 3. *Clutter and Lack of Whitespace:* Elements are placed very close together, especially the boxes on the right and the "Explore" tags at the bottom. Modern design uses whitespace effectively to separate elements and improve readability. 4. *Dated Typography:* The fonts and the way they are used (varying sizes, weights, mixing styles) don't feel contemporary.

*How to Make it Better in a Modern Style:*

The goal is to create an interface that is clean, intuitive, easy to read, and visually appealing, guiding the user naturally through the information.

1. *Simplify the Layout:* * *Reduce the number of distinct boxes:* Can some of the information be combined or presented more subtly? * *Use consistent containers:* If you need boxes or cards, make them look the same – consistent borders, shadows (if any), and background. * *Prioritize information:* What's the most important thing a user should see (probably the location and the Street View)? Make that prominent. Group related secondary information logically.

2. *Establish a Consistent Color Palette:* * Choose 2-3 primary brand colors, plus neutrals (whites, grays, blacks). * Use color meaningfully and sparingly, not just to differentiate every piece of content. For example, maybe use color for buttons, highlights, and key interactive elements, and use neutrals for background containers.

3. *Choose Modern Typography and Use it Consistently:* * Select one or two clean, modern fonts that are easy to read on screen. * Use a clear system for font sizes and weights (e.g., large bold for titles, standard for body text, smaller for captions). This creates a strong visual hierarchy.

4. *Use Whitespace Effectively:* * Add more padding inside elements and margins between them. This gives the interface "room to breathe" and makes content easier to scan.

5. *Refine Interactions and Elements:* * Ensure buttons have a clear, modern style and hover states. * Icons should be consistent in style (e.g., all outline, all filled). * Consider subtle animations or transitions where appropriate (but don't overdo it).

In short: Focus on cleanliness, consistency, and clear hierarchy. Think about popular modern apps – they often use flat design or subtle depth, clean typography, minimal color palettes, and plenty of whitespace to make the content stand out.

Your idea is genuinely interesting, and bringing the UI up to a modern standard would make it so much more inviting and user-friendly!

Hope this feedback is helpful!

I see a set of large images where the mouse cursor indicates they are clickable, but clicking in Firefox with AdBlock enabled doesn't work. I clicked for a long time and realized you're making fun of me?

PS: just use https://github.com/justinribeiro/lite-youtube

.cursorrules and other files that are added as system instructions in Cursor are just files, meaning the agent can edit them easely - you can tell it from the chat to add something to a rule and it will do it.

You are also wrong that it cannot do this automatically: if you add to the system instruction to record all important decisions in .cursorrules, it will record them there. automatically

Also, maybe I'm just not looking hard enough, but I can't find an RSS feed for the changelog of Cursor.

If projects have code on GitHub, it's easy to follow their updates, but if they are closed projects that post changelogs on their website, it's difficult for me to find an RSS feed. Usually, in the site code (like with Cursor), the feed leads only to blog updates.

Thank you for the feedback! I fixed the CSS, reduced the amount of bg blur, added the 'Files to Prompt' category, and many of the mentioned projects. I also added a mention of Chatbot Arena as a place where you can always see which model better for now.

I'm not interested in searching for the Origin country (I'm sure that in 90% of cases it will be the USA, then China) and Funding model, as I'm more into programming and I'm interested in the usefulness and stability of the tools. If someone does such OSINT and sends me the information, then of course I will add it.

I completely agree that we don't have well-established terms for categorizing all these tools. I really like your visualization by year of appearance.

Oh yeah, I've seen MCP popping up everywhere in AI themes. I haven't tried it out yet, but I've got a few tutorials (for Cursor and Cline) lined up. I'm keen to learn how it works and figure out if it's a good fit for us.

Yes, there is a semi-transparent background overlay. It works fine on my devices, but I will definitely investigate this issue and optimize the CSS.

UI/UX designing tools is exactly the category I am going to explore next, so yes, this section is not finished yet.

Why doesn't your MCP catalog contain a server for working with cursor.directory?

Why not in a three-dimensional space?