It's on Lovable so you can just fork it and take a look (the prompt is in supabase/functions/transform-render/index.ts):
Transform this idealized architectural rendering into the most brutally realistic, depressing photograph possible. This is the WORST CASE scenario - what the building will actually look like in reality:
- Set on a dreary, grey, overcast late November day with flat, lifeless lighting
- The sky is a uniform dirty grey, threatening rain
- All trees are completely bare - just skeletal branches against the grey sky
- The landscaping is dead, muddy, or non-existent. No lush gardens, just patchy brown grass and bare dirt
- Remove ALL people, the scene should feel empty and abandoned
- Any water features should look stagnant and grey
- Add realistic weathering, dirt streaks, and construction residue on the building
- The building materials should look how they actually appear, not the idealized clean version
- Include visible utility boxes, drainage grates, and other mundane infrastructure usually hidden in renders
- The overall mood should be bleak but realistic - this is what buyers will actually see on a random Tuesday in late autumn
- Maintain the exact building, angle, and composition, just strip away all the marketing polish
The goal is honest truth, not beauty. Show what the architect's client will actually see when they visit the site.
- American Giant is pretty good for their pullover hoodies. They'll wear out at the cuffs first, but I've kept a single hoody in use for like five years with some repair stitching.
- Standard Issue makes good waffle knit shirts. They'll last a few years depending on how often you wash them.
- Duluth Trading makes some good cotton shirts and boxers. Quality has declined slightly, but they're the best plain cotton shirts and boxers I've found so far.
- Big John makes denim jeans on old Levi looms. They even use cotton stitching.
- Carhartt makes some okay dressy dungarees. Their work pants are worthless these days though (in my experience). They've been pivoting to lifestyle for a few years now.
- Filson in my opinion has declined, but they're still pretty good. The socks are great, but they're overpriced.
(Only posting this because I've struggled finding decent clothes myself and it's hard to tell what's good when you're shopping online)
SOTA for frame interpretation today is probably RIFE (https://github.com/hzwer/ECCV2022-RIFE) as far as I know, which is fast as hell as well, and really good results still. But it's already 4 years old now, anyone know if there is anything better than RIFE for this sort of stuff today?
I'm building something like Lua for robots, you might want to check it out if you're looking to collaborate. I didn't know about Lua when I started it, but I did end up at an "everything is a table" metaphor because it seemed good for robotics. This does allow for cool things like hot reloading and such.
Although, we've since moved to having several distinct data structures which conceptually map to tables, but implementation and syntax-wise have differences (mostly for performance).
BTW Basis was a good idea, I remember reading about Nondeterministic replay is a big problem on platforms like ROS.
Yeah, this is where I landed a while ago. What problem am I _really_ trying to solve?
For some people it's an ideological one--we don't want AI vacuuming up all of our content. For those, "is this an AI user?" is a useful question to answer. However it's a hard one.
For many the problem is simply "there are a class of users that are putting way too much load on the system and it's causing problems". Initially I was playing wack-a-mole with this and dealing with alerts firing on a regular basis because of Meta crawling our site very aggressively, not backing off when errors were returned, etc.
I looked at rate limiting but the work involved in distributed rate limiting versus the number of offenders involved made the effort look a little silly, so I moved towards a "nuke it from orbit" strategy:
Requests are bucketed by class C subnet (31.13.80.36 -> 31.13.80.x) and request rate is tracked over 30 minute windows. If the request rate over that window exceeds a very generous threshold I've only seen a few very obvious and poorly behaved crawlers exceed it fires an alert.
The alert kicks off a flow where we look up the ASN covering every IP in that range, look up every range associated with those ASNs, and throw an alert in Slack with a big red "Block" button attached. When approved, the entire ASN is blocked at the edge.
It's never triggered on anything we weren't willing to block (e.g., a local consumer ISP). We've dropped a handful of foreign providers, some "budget" VPS providers, some more reputable cloud providers, and Facebook. It didn't take long before the alerts stopped--both for high request rates and our application monitoring seeing excessive loads.
If anyone's interested in trying to implement something similar, there's a regularly updated database of ASN <-> IP ranges announced here: https://github.com/ipverse/asn-ip
The YouTuber Will Prowse has an excellent site where he tracks his most recommended batteries (and other equipment like inverters) at any time. The prices are always changing, and there are new products all the time so check on the his list any time you are looking to buy:
Like the other commenter said, batteries are a lot cheaper if you are willing to shop around. His top recommended budget battery today is a 4x your Anker Solix's capacity, and around 1/4th the price. You can find many 5kWh server rack batteries for under $1000 now.
I would buy a dev board + build it yourself, you will get a much better experience then trying to reuse the existing thing.
I have written implementations target at specific boards. So go and buy one of these and boom stick it in anything you want. I have done this for my kids and have a bunch of different characters. My favorite is my daughter has a toy that pretends to be 'the ocean' it is so funny and existential.
I really loved the Sonatino[0], but can't get it anymore :(
If you start building something shoot me an email and would love to help! I want to unblock/enable this space so bad, I think these kinds of projects are just so delightful :)
Honestly we were looking at hatchet / pickaxe - similar vein of a project but more dev focused, but in the end realised our use cases were not all that complex so just built everything in a bespoke manner.
We used n8n for two things mostly - AI agents and process automation.
For AI we just built own MCP servers, and then the agents are quite easy to use as the major frameworks kinda help you with it. N8N’s AI is kinda just a UI layer for langchain - though we just used google’s adk.
For process automation - well there is so much options it’s not even funny.
I'm working on Typegres, a new data layer for the modern stack (TypeScript + PostgreSQL).
My take is that for years, ORMs have hidden the power of PostgreSQL behind generic, database-agnostic abstractions. This made sense in 2010, but now it's a bottleneck.
Typegres rejects this. It's a "translator, not an abstraction," designed to express the full power of PostgreSQL (all statements, built-in functions, etc.) in a type-safe TypeScript API.
The latest killer feature my take of "object-relational mapping done right": class-based models with methods that are actually composable SQL expressions. This lets you extend your tables with expressive logic and fully-composable relations.
I don't want to motivate anyone to switch back to Windows (because Microsoft), but for anyone who doesn't want to or can't switch, but also doesn't want to endure Microsoft's chosen way of entirely ignoring user needs but instead focusing on squeezing money out of them through ads and the collection of user data in Windows 10 and 11, check out the Windows 10 IoT Enterprise LTSC version! It's still supported 'til 2031 iirc. Plus, as it's meant to Enterprises and IoT systems, it's stripped from all the ads, bloatware, and what not. When I installed it first one year ago, I was just sad that I didn't know it earlier. It's everything I ever wanted Windows to be: lean, fast, and (somewhat) minimalistic, at least compared to stock Windows 10.
Can't 100% say whether Windows 11 IoT LTSC is equally good, but from what I've read it also is worth considering.
Talking of cheap and powerful devices one can also look at Chinese UZ801 4G LTE (Qualcomm MSM8916) dongles. They cost like only $4-5 and pack quite impressive HW: 4GB eMMC, 512MB RAM, actual 4G modem sometimes with 2 sim switching support. Since it's actually old Android SOC there is even GPU and GPS in there. And a lot of work was already done on supporting them:
Related: I know that many people use AI image generators to make pixel art, and recently I've stumbled upon a great tool to make a proper pixelart based on AI generated input see https://github.com/jenissimo/unfake.js and live demo on https://jenissimo.itch.io/unfaker
(disclaimer: I don't know the author, just thought I'd share as I find it amazing)
Speaking of operationalizing the woke mind virus, here's a funny story about GROK, Jessica Rabbit, Vivian Jenna Wilson, Marshal McLuhan, Jean-Paul Sartre, Hunter S Thompson, HAL 9000, and other simulated characters appearing together on a fictional episode of Mike Myer's classic SNL skit "Coffee Talk with Linda Richman":
Episode: "The MechaHitler Breakdown" - July 9, 2025:
If you're not familiar with the "Coffee Talk with Linda Richman" cultural reference, here is the most awesome epic episode (where Madonna disses herself, and a very special cameo appearance):
As you might guess it's kind of like The Sims meets Minsky's Society of Mind running in Cursor, and it includes simulated characters like Linus Torvalds who is great at practical stuff like helping out with git, devops, programing, and extremely harsh code reviews.
Simulated Linus's best burn (and possibly patentable great idea) was:
"I bet you even had pie menus in the bathroom! 'Which direction to flush?'"
But back to reality:
What was so shocking was the amount of energy Jonathan Payne puts into his hatred of trans people, constantly scanning his Facebook feed with paranoid zeal, relentlessly dropping in uninvited out of the blue on two different occasions to provoke fights in public with innocent people he doesn't even know, who were just trying to have a friendly discussion.
I hope I get too tired and lazy to hate that intensely and vigorously before I am so old that I lose my mind that badly.
Here is what I wrote to Arthur van Hoff and James Gosling about Jonathan Payne's behavior:
>I have many trans, gay, and otherwise marginalized friends, and Jonathan Payne posted a bunch of uninvited transphobic bullshit on my Facebook page in the middle of a conversation we were having about transphobia, which several of my trans friends were participating in, and where many others could see.
>This includes, among many other people, Lynn Conway, an 86 year-old trans woman, computer scientist, electrical engineer and transgender activist, who INVENTED superscalar computer architecture at IBM, who in 1968 FIRED her and destroyed her life for transitioning, then 52 years later in 2020 officially and publicly apologized to her. Later after putting her life back together from scratch under a new identity she worked at Xerox PARC and WROTE THE BOOK on VLSI design, and taught classes on VLSI design at MIT to none other than Guy Steele, who you certainly know, who in her class designed his famous hardware Scheme microprocessor, and to James Clark, founder of SGI and Netscape, who in her class designed his original 3d graphics accelerator hardware, and also to our own friend David Levitt.
>It was humiliating to me for Jonathan Payne to embarrass me in front of so many of my trans friends including Lynn Conway.
I have your problem. Philips makes no flicker LEDs, they don't have PWM like the Hue system, and good capacitors so no 60hz flicker. They're the "Ultra Definition Eye Comfort" models.
at Postman, we have a network for companies to reach developers and distribute their APIs on our platform. worth checking out: https://www.postman.com/explore.
you need to handle monetization, but the network gives you eyeballs
I’m building ninja.ai — it looks like a one-click App Store for MCP Servers, but the real goal is much bigger: creating a “Universal Fabric of Context” that lets AI tools tap into structured information across the web easily.
It started when I found it surprisingly hard for my partner to install and connect MCP Servers — even simple ones. I realised if we want AI agents to really interact with the web, it needs to be as easy as installing an app.
Right now, you can browse, install, and connect servers in one click. Over time, it’ll make AI integrations as easy as installing an app — no messy APIs, no custom scraping.
If you’re working with AI models, agents, or data-heavy tools, I’d love to hear what kinds of “context pipes” you’d want to see added.
The 8x16 font from the Atari ST's hi-res mode is pretty slick if you like something bold and a little futuristic.
https://github.com/ntwk/atarist-font (or rip it directly from the ROM)
And this is why I'm using Zed today. I'm deadly serious. I was a huge proponent of VSCode at first but I've soured on it, and now I don't want my workflow to depend on it in any way.
Awesome software, but I don't trust the upstream org further than I must.
Hey! The weakest part of Claude Code I think is that it's closed source and locked to Claude models only. If you are looking for inspiration, Roo is the the best tool atm. It offers far more interesting capabilities. Just to name some - user defines modes, the built in debug mode is great for debugging, architecture mode. You can, for example, ask it to summarize some part of the running task and start a new task with fresh context. And, unlike in Claude Code, in Roo the LLM will actually follow your custom instructions (seriously, guys, that Claude.md is absolutely useless)! The only drawback of Roo, in my opinion, is that it is NOT a cli.
I am a robotics engineer/scientist and I do shit ton of visualization of all kind of high-fidelity/high-rate data, often in a streaming setting - time series at a few thousand Hz, RGB/depth images from multiple cameras, debugging my models by visualizing many layer outputs, every augmentation, etc.
For a long time, I had my own observability suite - a messy library of python scripts that I use for visualizing data. I replaced all of them with rerun (https://rerun.io/) and if you are someone who think Scipton is exciting, you should def try rerun too!
I use cursor/vscode for my development and add a line or two to my usual workflows in python, and rerun pops up in it's own window. It's a simple pip installable library, and just works. It's open source, and the founders run a very active forum too.
Edit: One slightly related tid-bit that might be interesting to HN folks. rerun isn't that old, and is in active development, with some breaking changes and new features that come up every month. And it means that LLM are pretty bad at rerun code gen, beyond the simple boilerplate. Recently, it kind of made my life hell as all of my interns refuse to use docs and try using LLMs for rerun code generation and come to me with a messy code spaghetti. It's both sad and hilarious. To make my life easier, I asked rerun folks to create and host machine readable docs somewhere and they never got to it. So I just scrape their docs into a markdown file and ask my interns to paste the docs in their prompt before they query LLMs and it works like a charm now.
Transform this idealized architectural rendering into the most brutally realistic, depressing photograph possible. This is the WORST CASE scenario - what the building will actually look like in reality:
- Set on a dreary, grey, overcast late November day with flat, lifeless lighting - The sky is a uniform dirty grey, threatening rain - All trees are completely bare - just skeletal branches against the grey sky - The landscaping is dead, muddy, or non-existent. No lush gardens, just patchy brown grass and bare dirt - Remove ALL people, the scene should feel empty and abandoned - Any water features should look stagnant and grey - Add realistic weathering, dirt streaks, and construction residue on the building - The building materials should look how they actually appear, not the idealized clean version - Include visible utility boxes, drainage grates, and other mundane infrastructure usually hidden in renders - The overall mood should be bleak but realistic - this is what buyers will actually see on a random Tuesday in late autumn - Maintain the exact building, angle, and composition, just strip away all the marketing polish
The goal is honest truth, not beauty. Show what the architect's client will actually see when they visit the site.