For those of us programming nerds that want to play with aerodynamics, I can't recommend AeroSandbox enough. While the code is pretty obviously written for people who know their way around aerodynamics and not so much around programming, it is remarkably powerful. You can do all sorts of aerodynamic simulations and is coupled with optimization libraries that allow you to do incredible aerodynamic optimizations. It comes included with some pretty powerful open weight neural network models that can do very accurate estimates of aerodynamic characteristics of airfoils in a fraction of the time that top tier heuristic solvers (like xfoil) can do (which are already several orders of magnitude faster than CFD solvers).
Ok it might sound crazy but I actually got the best quality of code (completely ignoring that the cost is likely 10x more) by having a full “project team” using opencode with multiple sub agents which are all managed by a single Opus instance. I gave them the task to port a legacy Java server to C# .NET 10. 9 agents, 7-stage Kanban with isolated Git Worktrees.
Manager (Claude Opus 4.5): Global event loop that wakes up specific agents based on folder (Kanban) state.
Product Owner (Claude Opus 4.5): Strategy. Cuts scope creep
Scrum Master (Opus 4.5): Prioritizes backlog and assigns tickets to technical agents.
Architect (Sonnet 4.5): Design only. Writes specs/interfaces, never implementation.
Archaeologist (Grok-Free): Lazy-loaded. Only reads legacy Java decompilation when Architect hits a doc gap.
CAB (Opus 4.5): The Bouncer. Rejects features at Design phase (Gate 1) and Code phase (Gate 2).
Librarian (Gemini 2.5): Maintains "As-Built" docs and triggers sprint retrospectives.
You might ask yourself the question “isn’t this extremely unnecessary?” and the answer is most likely _yes_. But I never had this much fun watching AI agents at work (especially when CAB rejects implementations).
This was an early version of the process that the AI agents are following (I didn’t update it since it was only for me anyway): https://imgur.com/a/rdEBU5I
> AI also really has trouble with transcribing my speech. I noticed that as early as the '90s with early speech recognition software. It was completely unusable.
I don't know what your transcription use cases are, but you may be able to get an improvement by fine-tuning Whisper. This would require about $4 in training costs[1], and a dataset with 5-10 hours of your labeled (transcribed) speech, which may be the bigger hurdle[2].
1. 2000 steps took me 6 hours on an A100 on Collab, fine-tuning openai/whisper-large-v3 on 12 hours of data. I can shar my notebook/script with you if you'd like.
2. I am working on a PWA that makes it simple for humans to edit initial, automated transcriptions with mistakes for feeding the correct dataset back into the pipeline for fine-tuning, but its not ready yet
Not for takeoff. Apollo astronauts used barbituates to help sleep, and scopolamine with dextroamphetamine during re-entry. Unsurprisingly the average HR during re-entry was rather higher than at launch! (source: https://space.stackexchange.com/a/31987 )
Gemini would likely have been similar, save for using Cyclizine during re-entry rather than scopolamine/dextroamphetamine.
Receiving 433Mhz sensor data using rtl_433[0] with an RTL SDR was a lot of fun when I started doing it last year. There's MQTT output if you want to send it to Home Assistant, et. al., as well as simple text output to stdout. It was great fun seeing my neighbors' sensors, tire pressure sensors in passing vehicles, etc.
There a ton of devices that use 433Mhz. You can also extend rtl_433 pretty easily.
Simon Tatham's Puzzles app has a minesweeper version that both A) generates on first click so you are guaranteed to never hit a mine on the first click and B) every board is 100% solvable with no need to guess. I have no idea what method it uses. I believe that _every_ puzzle in it is 100% solvable with no need to guess. It's also add free and completely local. It's a great app for anyone who likes puzzles on their phone.
Allow me to give you a different viewpoint. And this is coming from someone that has an _amazing instinct_ to be in the "Who The Fuck Cares" club. I use that instinct to protect my mental health but nothing more than that.
What I noticed when I checked out at work is that it also makes me check out in my personal life (PL). It bleeds in. Generally, in my personal life I'm not checked out. That bleeds into work.
So work bleeds into PL and PL into work. I found that it was painful for work to bleed into my PL like that since I'm switched on and I just had this hint of "ah... whatever who gives a fuck."
I give a fuck.
I give a fuck because it's my life. I do it for myself. I don't do it for my boss or my colleagues. I do it for me.
I've found that this attitude is way more helpful to me as two things happen:
1. I'm more productive at work so I don't have to cover my ass at all. When I was in the "Who The Fuck Cares" club, I needed to cover my ass once per month (read: I didn't do anything for like 3 days and people were expecting results on day 4).
2. Since it's in service for my personal life, I don't go too far. The moment I notice that work encroaches too much upon personal life, my instinct comes back immediately and I pay my visit to the "Who The Fuck Cares" club, and party as long as I want to.
This is exactly why I still play Go, practice martial arts, archery and use the command line for my dev workflow. Those are all arguably less efficient and obsolete. The AlphaGo series can defeat the strongest human Go players and firearms are more effective than unarmed martial arts and archery. GUI is easier for most people than the command line. Yet, I practice all these to develop my mind and body. I don't have to be a world-class Go player to benefit from learning to play Go.
This happens a lot in the natural world in ecosystems. For example, many people plant trees and add a drip system. The trees grow to depend on the drip system, and never stretch and develop their roots -- and the relationship they have with the soil microbiome. They make the trees prone to get knocked down when an unusually strong gust of wind come through.
> If you’re repeatedly drawn to a thought, feeling, or belief, write it out. Be fast, be sloppy.
I couldn't agree more here! A friend has wanted to start a writing/journaling habit for a long time, but didn't know what to write about. I told him, don't think to write—write to think [1].
Show up to an empty page, without knowing, is totally acceptable! So is writing things down that make you feel embarrassed, confused, etc.
When I'm journaling, I often find prompts/frameworks helpful for guiding this escape.
I really like Byron Katie's framework, which she calls The Work [2]. After you notice and draw to mind a stressful thought, answer these four questions:
Q1. Is it true?
Q2. Can you absolutely know that it’s true?
Q3. How do you react, what happens, when you believe that thought?
Q4. Who would you be without that thought?
Then, invert the thought. She writes, "Turn the thought around. Is the opposite as true as or truer than the original thought?"
Derek Sivers also shares some really great questions for journaling for reframing [3].
This is the key to a lot of my workflows as well. I'll usually tack some form of "ask me up to 5 questions to improve your understanding of what I'm trying to do here" onto the end of my initial messages. Over time I've noticed patterns in information I tend to leave out which has helped me improve my initial prompts, plus it often gets me thinking about aspects I hadn't considered yet.
The first part of this, where you told it to ask YOU questions, rather than laboriously building prompts and context yourself was the magic ticket for me. And I doubt I would have stumbled on that sorta inverse logic on my own. Really great write up!
I tend to believe that whatever a person ends up doing is what they actually "want" to do at some inner emotional level (which they may not consciously be aware of).
So this person wants to build, but after releasing things to lukewarm feedback, wants to not work on them anymore. Why would they want that? Well, that is a pretty normal behavior I think. But I would speculate...
I wonder if this person is like me? Because they sound kinda like it. If so, I'd guess that they know that they're capable of a lot, so they're inspired to work really hard at goals, but that they're also mortified at presenting themselves to the world as confident and capable publicly. So they can work really hard when it's private and based on their internal self-image, but when it becomes public they have to reconcile it with their public embarrassment at who they are, and so they choose (in the sense of: their fear chooses for them) instead to hide and go back to working in private where it is safe. They are hoping for something else to happen: probably, they release one of their projects and it is so well-received, and garners so much respect, that they find a new public identity which they can step into and inhabit, allowing them to be proud of themselves and solving their inner problem.
I wonder (more analogy to myself): was this person bullied or ostracized as a child, and made to feel like being confident and proud in public was shameful? If I've guessed right I hope you figure it out (and then please tell me what you find cause I could use some tips...). I wonder what they want to have happen when they release a project? What, if they had to admit to it, would their most happy outcome upon releasing a product look like? Probably it will be sort of embarrassing to even describe, but the details are probably telling: they will point straight at the core anxiety. It will be a fantasy about the anxiety being cured.
Just speculating. But I would imagine that the explanation here isn't the real one, because the actual explanation will cause their behavior to completely change, instead of just building an intellectual edifice around it. (Personally I am very skeptical of any explanation that goes "because ADHD", because, what causes the ADHD? In some people it seems to be intrinsic but in lots of others it's clearly induced by something social. So is it fixable? Are there stories of it being "fixed" by solving the social issue? I dunno. But surely the goal is to somehow not live in this tension anymore: either fix the thing or let it go, and not be stuck in a loop of trying to get it fixed by some external blessing.)
I'm kidding. Once you know what, you can check your patterns against the standard and the deviations. Then you can approach the circuits that are fucked and identify triggers.
Best to get to the roots of the reinforcement in childhood, though.
Behaviors and preferences surpressed, people you didn't beat up for reasons and so on, behavioral triggers, when you held back and why. What distracted you when you did a specific thing or the other. And, of course, the standard stuff, what were you rewarded for and did you really care about that thing yourself?
I was working some time ago on a 4X with an hexagonal map (not only made of hex tiles, but the overall shape was an hexagon rather than a rectangle, no wrap-around).
Using a noise function alone does not let you choose how many landmass you want to generate.
Using "wave function collapse"/constraint solvers is too slow for large maps when the amount of constraints increase (please don't put a mountain in the middle of the ocean, or a snow tile in the middle of a desert). My implementation took almost 8h to generate a single map, and the result was not even good.
In the end, I used a combination of multiple techniques:
- voronoi to split the map into regions
- use a noise function to make the regions boundaries a bit more natural
- fill the map with water tiles, place randomly some island seeds
- grow the islands from their seeds using a cellular automata
- to create continents, simply put a lot of island seeds in the same area, to generate a bigger one
- place mountains or rifts on region boundaries to simulate "tectonic plates"
- generate a heat map (influenced by position of the north/south poles and equator of the map), a humidity map (influenced by leftover ocean tiles), a height map (which is influenced by the already placed mountains/rifts) using a noise function
- using the previous heat/humidity/height maps, generate a wind map
- using the wind map, modify the humidity map (wind carries humidity over the land)
Then, choose randomly some tiles that fit some criteria to place biomes:
- desert on hot/dry land
- forest on temperate land
- swamp on temperate/wet land
- jungle on hot/wet land
- ...
Then a bunch of different cellular automata to grow the biomes naturally.
The result was quite nice, but it still wasn't on par to the map gen in Civilization games. I still want to continue this project, but I think in my heart I gave up.
EDIT: Some formatting because i always forget HN does not support markdown lists
Actually, "git reset --soft" moves the current branch to another commit, without moving the index (aka staging area) along with it, whereas "git reset" (aka "git reset --mixed") moves the current branch AND the index to another commit. I really couldn't wrap my head around it before I had gone through "Reset demystified" [1] a couple times - it's not a quick read but I can strongly recommend it.
Google Calendar supports CalDAV, but only behind OAuth, which many clients do not support. I've created https://github.com/bjesus/oauth-hopper to solve that - it takes care of the OAuth steps and provides a clean CalDAV endpoint that you can use to read and write to your calendar from almost any calendar application. In reality OAuth Hopper can be used to abstract OAuth away from any endpoint - it isn't CalDAV specific in anyway.
> copy and paste content and not having layout page is annoying at times
HTML was envisioned as an SGML application/vocabulary, and SGML has those power features, such as type-checked shared fragments/text macros (entities, possibly with parameters), safe third-party content transclusion, markup stream processing and filtering for generating a table of content for page or site navigation, content screening for removal/rejection of undesired script in user content, expansion of custom Wiki syntax such as markdown into HTML, producing "views" for RSS or search result pages in pipelines, etc. etc. See [1] for a basic tutorial.
Then try your own designs combining properties of several designs. Start with simple blinkers and buzzers at first... Try to avoid Arduino designs until you've done a few 555, transistor, and opamp circuits first. =3
This post deserves more attention, I think. It's occurred to me as well.
Over the holidays, my father gave my children a book that he had written. It was a photo essay that was 50 pages, and it was titled 'Sharks'. It's an unpublished labor of love that he spent about 500 hours on.
It's a true story centered on Captain Frank Mundus, who operated the Cricket II. He was a renowned shark fisherman and would take people out to fish for enormous sharks. He did this for 40 or 50 years.
An author by the name of Peter Benchley wrote a novel that was heavily inspired by many of Frank's traits, his mannerisms, his approach to shark fishing, the kind of boat he had, the kind of charters he ran. The novel was titled 'Jaws' and received little attention when it was first released. A while after, a director by the name of Steven Spielberg took notice of it and turned it into a multi-million dollar blockbuster movie.
My father was a lawyer that Frank Mundus consulted with and asked, is there any way that he could get a payout for being the inspiration for this character?
My family read the book over the holidays, and it was clearly my father's position that Steven Spielberg and Peter Benchley were maybe the sharks that the title of the book was talking about. The idea that they could make $100 million based on the work and life of this captain and give him literally nothing in return, not even attribution, seemed wrong to him.
I was the lone detractor in the room. My take is that Captain Frank Mundus was just living his life. He was doing what he did to make money chartering fishing trips for sharks. He would have done this regardless of whether or not a writer had come along or a movie had come along. What Peter Benchley and Steven Spielberg did is they found value in his work that he didn't know existed and that he wasn't capable of extracting. I think this is generally true of artists. They wander the world and they create art that gives the viewer a new insight into the experiences the artist had. If artists had to give money back to every real-life inspiration, I think the whole system wouldn't work.
I see parallels with the current attitudes toward AI. I think writers are a lot like Captain Mundus. They're living their life, they're writing their stories, or doing their research and publishing, and having people read their works. And copyright is helping them do all this.
AI companies have come along and found value in their work that they didn't know existed and they were never capable of extracting. And that's OK: that's what innovation is, taking the work that others have done and building on it to create something new.
I'm not unequivocally in favor of all applications of AI, but I do think there are tons of places that can be super helpful and we should allow it to be helpful. One example: I'm drafting this on my phone using Futo keyboard entirely with my voice. Extremely useful, but no doubt trained on copyrighted content.
Current embeddings are badly trained and are massively holding back networks. A core issue is something I call 'token drag'. Low frequency tokens, when they finally come up, drag the model back towards an earlier state causing a lot of lost training. This leads to the first few layers of a model effectively being dedicated to just being a buffer to the bad embeddings feeding the model. Luckily fixing this is actually really easy. Creating a sacrificial two layer network to predict embeddings in training (and then just calculating the embeddings once for prod inference) gives a massive boost to training. To see this in action check out the unified embeddings in this project: https://github.com/jmward01/lmplay
This is very cool. BTW, when developing single HTML file apps, instead of localStorage, one can use the HTML as the source of truth, so the user can just save/save-as to persist. I had mentioned my quick and dirty attempt at an image gallery that is a self-contained html file and some really liked the concept, if not the "app" itself: https://news.ycombinator.com/item?id=41877482
The Telharmonium dated from the "if only we had gain" era of pre-electronics. The thing was a huge collection of sizable AC generators running at different
frequencies, run through a keyboard, and mixed with transformers. With no way to amplify a small signal, there was no way to downsize the thing. Once amps were invented, the Hammond Organ, with its tone wheels, was the same concept in a piano-sized package.
For those of us programming nerds that want to play with aerodynamics, I can't recommend AeroSandbox enough. While the code is pretty obviously written for people who know their way around aerodynamics and not so much around programming, it is remarkably powerful. You can do all sorts of aerodynamic simulations and is coupled with optimization libraries that allow you to do incredible aerodynamic optimizations. It comes included with some pretty powerful open weight neural network models that can do very accurate estimates of aerodynamic characteristics of airfoils in a fraction of the time that top tier heuristic solvers (like xfoil) can do (which are already several orders of magnitude faster than CFD solvers).
https://github.com/peterdsharpe/AeroSandbox