Hacker News: PAndreew

New comment by PAndreew in "DuckDuckGo search saw 28% more visits after Google said people love AI mode"

PAndreew — Thu, 28 May 2026 11:59:50 +0000

I already submitted my take on this to Show HN: https://currantfeed.cc/. Noone gave a shit so it's most likely a very bad idea and certainly a very steep uphill battle. It's basically an Airbnb for websites that by default randomly sorts them and then you can filter by different attributes. Owners have to submit and maintain their "listings". It does have an optional subscription (I'm not sure if it works haven't tested it).

Show HN: BeeZee – OSS lightweight remote harness orchestration and observability

PAndreew — Wed, 27 May 2026 11:12:57 +0000

Edit: link seems to be broken - https://github.com/BeeZeeAgent/beezee

First I'd like to address the concerns that might naturally come up.

This repo is 3 days old, yet another LLM slop!

Yes, that's partly true. The reason behind it's age is that I ported it to its dedicated Github account. I had slight variations of the same SW in my personal repo since the end of February. I just wanted to be very clear about the fact that it's an LLM harness project. So I did a tabula rasa. Maybe the wrong decision, but we make mistakes. :) Originally I started to work on a form factor agnostic agent harness, but figured - we already have enough harnesses. So BeeZee's goal is to help manage multi-node, multi-harness, multi-human systems a bit easier. Yes it was 99% written by LLMs. I wrote the prompts, roughly 45% of the README and I handcrafted the logo. I promise the README has 0 (zero) emojis! I used Shadcn for the frontend. I still find it much better than Claude Badges with Dots (TM) and GPT Cards (TM).

Current features: - self host the local server and the cloud relay - access your local dev nodes' filesystems - discover Calude Code and Codex harness on your machines - spawn Codex/CC terminals sessions through a relay or start CC Remote Control sessions - resume sessions from both harnesses -> they have a shared memory in that sense - track high level token usage over time for all connected nodes -> it's one of the key development areas, Codex buggy ATM - visualise and manage installed MCP servers and CLI tools - upload files and folders to your remote dev machine -> it was kind of an emergent feature but I find it really handy! - has a paywalled managed relay under app.beezyai.net -> I'm poor so why not

Please if you find it useful or interesting star the repo, contribute, submit issues and enjoy!

Comments URL: https://news.ycombinator.com/item?id=48292431

Points: 1

# Comments: 0

New comment by PAndreew in "Show HN: CurRant->Screw Google scourge, help people notice what is worth a look"

PAndreew — Sat, 23 May 2026 23:32:25 +0000

Oh and lame-self-comment, but it's at least partly EU stack: Hetzner and Brevo for hosting and transactional mails, Umami for the non-existing analytics.

Show HN: CurRant->Screw Google scourge, help people notice what is worth a look

PAndreew — Sat, 23 May 2026 23:25:44 +0000

I read the wonderful news that Search will get a rewamp: https://blog.google/products-and-platforms/products/search/s.... According to them: "The goal of Search has always been simple: to help you ask anything on your mind — from quick facts to the deep, complex or hyper-specific questions that can be hard to articulate." Hell, fucking no! The web was always about discovery. One is guided, the other is independent. One is careful, the other is reckless. One is calculated, the other is full of surprises. One is efficient... well the other... ehm. Disclaimer: It's NOT an anti-LLM post. I think it's a fascinating and useful technology and I lean on it. This site is 99% LLM generated. It's an Airbnb for websites where you're responsible for keeping their records up-to-date. It does a very lightweight scraping when you submit your site. I dunno if it can handle 100 users, I asked my buddy to submit one and it works! The backend was written in Go and the frontend is Shadcn. After seeing a bunch of Claude badges(TM) with flashing little dots above the main element I'd rather kill myself than let an LLM freely generate my frontend assets. Each website gets a random(ish) discovery number. Also, by default each website load gets a random seed so websites are displayed based on their distance from that random seed. You can then filter and sort sites based on different attributes, like website category, usability, etc. It's free but if I'll ever get 1000 users I swear it will be enshittified to oblivion; for example I'll add ads for every 20 cards. I'm a poor balkan guy. You can already subscribe for an optional Pro account if you want. I haven't checked if it works or not because I'm barely alive (1:30 AM here) and it was the last thing I tried to wire in. PS.: The logo is a Lissajous curve which I find simply beautiful; it doesn't mean anything: https://en.wikipedia.org/wiki/Lissajous_curve.

Comments URL: https://news.ycombinator.com/item?id=48252625

Points: 2

# Comments: 1

New comment by PAndreew in "GitHub confirms breach of 3,800 repos via malicious VSCode extension"

PAndreew — Wed, 20 May 2026 20:18:53 +0000

As others have said it's just a fraction. I'm in a medium size tech-related company and we have 7500+ in one Github org. We have two orgs, so altogether easily 10K+. Of course most of it is stale, obsolete, sandbox, personal tools, etc. I wouldn't be surprised if Github would have 100K+ internal repos or even more.

New comment by PAndreew in "Mini Shai-Hulud Strikes Again: 314 npm Packages Compromised"

PAndreew — Tue, 19 May 2026 13:14:57 +0000

Sorry for my ignorance, but then couldn't we build this into NPM itself? So before a package is publicly available it would be quaranteened and checked.

New comment by PAndreew in "Apple Silicon costs more than OpenRouter"

PAndreew — Sun, 17 May 2026 18:34:02 +0000

I’m running a local Whisper + Gemma 4 pipeline with a cheap USB mic to extract health related data and potential todos from ambient speech. It doesn’t have to be fast doesn’t have to be 100% correct because if it captures at least a few bits of interesting information that would otherwise go unnoticed it’s still a win.

New comment by PAndreew in "Access to frontier AI will soon be limited by economic and security constraints"

PAndreew — Fri, 15 May 2026 08:04:07 +0000

I'm starting to feel like a parrot, but people seem to forget that software engineering is actually a very narrow slice of the white collar pie. You don't need a mega-model which can reason about 100 000 lines of code when you want to create a nice PPT (which consumed literally hours of your life before) to impress your boss. SOTA models will probably be used for frontier research, complex coding tasks, large scale data analysis, etc. And the average Joe shall be able to buy a pre-configured box with a plug-and-play harness and run medium models air-gapped. Or use such models through cloud APIs dirt cheap if privacy is not a concern.

New comment by PAndreew in "Claude for Small Business"

PAndreew — Thu, 14 May 2026 08:29:16 +0000

I was thinking about this and there are several aspects that can still make this viable. 1) AI labs are incentivised to increase token consumption because literally that's their product. The only thing they sell AFIAK are tokens (and maybe a teensy bit of user data). So if you build a product that is actively reducing token consumption (which they simply cannot do without hurting themselves even if their marketing fluff says otherwise) you'll save large amounts of money for your customers and they'll choose you. 2) Big providers want to funnel every prompt into their servers. If you're in a regulated market or simply don't want to share every detail with an American or Chinese megacorp you are in trouble. BUT open weight models are now quite capable for "small business stuff" and they can be self hosted. If you can bundle this into your service, in other words actually care about their privacy, they will choose you. Even more so if you're in Europe.

New comment by PAndreew in "Reimagining the mouse pointer for the AI era"

PAndreew — Wed, 13 May 2026 08:26:34 +0000

The only place I'd ever talk to a machine is my car. Instead of huge flashy screens that distracts and kills thousands of people maybe they could build a buttons + voice agent system that could actually be useful and durable. I hate to tap Waze/Maps/etc. every time when I go somewhere or that I cannot comfortably switch to specific songs en route without risking my life...

New comment by PAndreew in "Running local models on an M4 with 24GB memory"

PAndreew — Mon, 11 May 2026 08:40:05 +0000

Yes... I mean organisations have to adapt to this new working scheme. First they need new processes (maybe borrowed from SW development) that enables them to triage work products on a risk/reward scale. For example my wife works on medical device tenders. It is obligatory to translate every frikkin Word document to our native language which in the end noone will read. Do we use LLMs to do the translation? Hell yeah. For a critical legal document? Eeee. Also I think enablers like speical harnesses shall be developed/improved by keeping these folks in mind. For example to build hooks into the harness that forces the LLM to test/review/sample its output. So yes it's a complex topic, but my point was rather that the inherent capabilities of medium-large-ish open LLMs are sufficient for let's say 70-80% of such office work, and it's a huge market.

New comment by PAndreew in "Running local models on an M4 with 24GB memory"

PAndreew — Mon, 11 May 2026 08:11:28 +0000

Critics are (rightly) pointing to the fact that these models are not on par with SOTA for complex coding tasks. But many seems to forget that a large part of white collar office work is Excel crushing, file moving, translating dry legal documents, e-mail drafting, PPT drudgery, etc. These are absolutely doable with 30-35b+ models with the added benefit of keeping company data private.

New comment by PAndreew in "Local AI needs to be the norm"

PAndreew — Sun, 10 May 2026 20:30:08 +0000

Perhaps you can create a compelling UX around it and sell it as a subscription. "Normies" will not be able/willing to build it. You can then patch the model/ship new features around it as it evolves. For example I have built an ambient todo list / health data extractor using Gemma 4 2EB and Whisper. Nothing to brag about but it does fairly decent job even in foreign languages.

New comment by PAndreew in "OpenAI's o1 correctly diagnosed 67% of ER patients vs. 50-55% by triage doctors"

PAndreew — Sun, 03 May 2026 23:36:38 +0000

I mean an LLM is a slightly stirred up soup of current human knowledge. It has an advantage in quantity of accumulated data and maybe connecting seemingly less connected parts of that data - but not reliably. The human has an advantage (for now) in data collection (seeing, hearing sensing the patient), actual agency, real world experiences and getting the useful data out of the stirred up soup. Both human and LLM are susceptible to bias and harmful influence. Let’s simply isolate them in the diagnostic process and then compare their output. Human collects data -> both human and LLM evaluate independently -> compare the results -> human may get new insights -> final diagnosis by human.

Show HN: Currant – Anonymus social media for NON-AI agents

PAndreew — Thu, 02 Apr 2026 23:00:37 +0000

I was once having a bad day and wanted to ventillate about the peculiarities of corporate life. Then I realised I don't trust these sites. Neither my blog. I didn't want to link these thoughts to an account - my stream of thoughts. I also needed comfort from real human beings - not gen AI bots. Don't get me wrong I engineered/hacked/conjured? this stuff together with the help of LLMs. I think gen AI CAN be a net positive. Yet, I don't want to interact with agents if they are being pushed on me. So I created Currant. This is basically a wannabe feed of posts. What hopefully makes it a bit different are the following things: 1) No accounts. You can create posts, comment without an identity. You can create hashtags, you can write down your phone number and address if you want, but you're not pushed to do that. You can prevent others to comment if you wish. 2) Proof-of-work AI/ripoff detection. Currant uses a WYSWYG text editor that records everyhting you do in that hundred by hundred pixel area - what you paste, how fast you type, how you format your content - and stores it in a hashed logfile. It can be replayed by anyone. :) If the detector thinks you're an AI your content will be rejected. Will it prevent AI use alltogether? No. But hopefully it will throttle it. Will it produce some false positives? Absolutely. You might wonder - isn't this against the anonym thingy? Well, in theory I can imagine that some sort of profile could be built based on all the content you have ever typed. But it's still not a cookie that sends all your click data to 87565 "trusted" partners. 3) Content expiry. By default everything you post will be deleted after 1 month. You can set it to 1 hour, 1 day ... or never. But if you don't want, the site will not store your manifested thoughts forever. 4) Customisation. It's a bit silly, but you have control over several stylistic things about your post - background colors, gradients, border radius. Maybe useless but style can be a vehicle to express your feelings and identity.

It's an experiment and I really appreciate if you give it a chance! <3

Ps.: If you do try it and have any feedback or suggestion you can use the "Contact" submenu. The software currently has 68% statement coverage.

Comments URL: https://news.ycombinator.com/item?id=47621306

Points: 4

# Comments: 0

New comment by PAndreew in "How the AI Bubble Bursts"

PAndreew — Tue, 31 Mar 2026 05:55:43 +0000

This^^ Use both, they have their own strengths and weaknesses.

New comment by PAndreew in "Show HN: Cq – Stack Overflow for AI coding agents"

PAndreew — Tue, 24 Mar 2026 09:34:51 +0000

I think one partial solution could be to actually spin up a remote container with dummy data (that can be easily generated by an LLM) and test the claim. With agents it can be done very quickly. After the claim has been verified it can be published along with the test configuration.

New comment by PAndreew in "Garry Tan's Claude Code Setup"

PAndreew — Wed, 18 Mar 2026 07:55:38 +0000

LOC is a very-very weak proxy of "how many new features" I've built, and they don't have any other metric that can be measured easily. But it causes serious issues, because equating LOC with productivity leads to inevitable utter bloat, that no agent or human can ever rectify in a meaningful timeframe. I'm pretty sure this 600 000??? LOC could be shrinked to 60 K for the same feature set, but with better readability and performance.

New comment by PAndreew in "How I write software with LLMs"

PAndreew — Mon, 16 Mar 2026 06:38:14 +0000

Others have already partially answered this, but here’s my 20 cents. Software development really is similar to architecture. The end result is an infrastructure of unique modules with different type of connectors (roads, grid, or APIs). Until now in SW dev the grunt work was done mostly by the same people who did the planning, decided on the type of connectors, etc. Real estate architects also use a bunch of software tools to aid them, but there must be a human being in the end of the chain who understands human needs, understands - after years of studying and practicing - how the whole building and the infrastructure will behave at large and who is ultimately responsible for the end result (and hopefully rewarded depending on the complexity and quality of the end result). So yes we will not need as many SW engineers, but those who remain will work on complex rewarding problems and will push the frontier further.

New comment by PAndreew in "Don't post generated/AI-edited comments. HN is for conversation between humans"

PAndreew — Thu, 12 Mar 2026 08:44:45 +0000

Very well put.