Yes, in fact my ability to draw from memory places I've seen is, if anything, from experience, substantially above average. However, when I draw from memory, my drawings do tend to be more stylized.
I can "imagine" what a building looks like, but I can't see it. Even our language makes this hard to describe, and the reason why I didn't realise that this isn't how other people do it, because most of our language for remember how something looks uses terms that implies we see them, so I just assumed until a few years ago (and I turn 50 this year) that that was just a metaphor to everyone else too.
How do you imagine (there's that word again...) blind people remember places they can't see?
To me, the notion that I'd need to "see" what the building looks like to draw it is bizarre because I know where everything is in relation to each other, and their shapes, so why would I need to see it?
EDIT: increasingly, and in part drive by split-brain experiments, I tend to think that a whole lot of what we see as our conscious decision-making, are instead shallow retroactive attempts by parts of the brain at rationalising decisions largely already taken autonomously into a cohesive self/ego.
Basically recreated LeetCode on a $5/month VPS lol. It started as a class project but over 800 people joined so I kept refining it for 2 months. It supports running Python, Java, and C++ code (building a code runner is tough). Give it a try and let me know what you think!
Writing on a blog is a very inexpensive way to establish your credibility about different subjects. This pays off later down the line when you can link people to things you've written in the past.
Credibility is a very valuable commodity. It's worth investing in ways to build more of it.
Don't assume people will stumble across your content (though they will eventually via Google). Actively send links to people who you are already engaged in conversation with.
It's not the number of readers you have that matters: it's their quality. I'll take a dozen people reading my stuff who might engage with me usefully or lead to future opportunities over a thousand readers who don't match that criteria.
This is pretty much the methodology behind "comprehensible input", where you consume lots of content that you can just about understand and "let your brain figure it out".
There's quite a lot along these lines. LingQ helps you learn as you read books, and I built https://nuenki.app, which gives you constant comprehensible input as you browse the web.
I also really like Language Transfer, which isn't really a comprehensible input course, but tries to draw parallels to English and talks through the etymology a little. The approach appeals to me.
Thanks for sharing your earlier comment—it’s an excellent idea! Combining spaced repetition with knowledge graphs could revolutionize how concepts are mastered, especially in subjects with hierarchical dependencies like math, science, or even music theory. I’m curious, though: do you think this approach works best for self-paced learning, or could it be integrated into a classroom setting? Also, how would you envision making it engaging for younger learners who may not naturally gravitate toward structured systems?
I'll definitely prepare a longer write-up when I have everything figure out, but here's a summary:
I have 4 systems:
- Komfovent HRV for ventilation
- NIBE F-series heat pump for floor and water heating
- Vaillant gas boiler that "supports" the heat pump
- Samsung multi-split AC units
HRV - Komfovent uses the same controllers in all of their units, so you get all the communication goodies you'd want - though it took me a long while to figure out that basic features need to be toggled on :) There are existing YAML presets for their C6/C6M controllers on HA forums. The only caveat is that if you want to feed it a virtual thermostat, you need a stuff a device simulating a 10k NTC inside of the ventilator. Otherwise it's just a single Ethernet cable.
Heat pump - I'm not exactly sure if I'm happy with NIBE, but thanks to the community the integration ended up being quite easy. I wasted a bunch of money on their MODBUS40 just to learn that you need to use a certain MODBUS address in the internal bus to make certain registries writeable (eg. thermostat values) - so I took an ESP32 with Ethernet, a galvanically isolated RS485 dongle, a 12V to 5V converter and used https://github.com/elupus/esphome-nibe. The firmware extracts my templated HA sensor's value and feeds it to the heat pump as a virtual thermostat.
Vaillant uses this weird "eBUS" protocol, there's a bunch of cheap PCBs that you can use to connect to it - I'm using https://github.com/danielkucera/esp-arduino-ebus. That's the last system that I haven't touched :)
Samsung ACs use their MIM-B19N modules installed in the outdoor units. There's some magic around enabling remote control, but once you plug their diagnostics device into their indoor units, you can flash all of them at once. I had to mess around with internal NASA addresses to have all the units appear at once.
For indoor sensors I have 3 types:
- AirGradient units measure CO2, tempeature, humidity, PMx etc. - these are mounted at ~150cm and feed the "current house temperature" template.
- I have like 8 Everything Presence One devices, powered by a custom PCB that converts 12V/24V sent over wired alarm cables to the device. They have built-in temperature, humidity and motion sensors. These are mostly installed for motion sensing and their height makes the temperature measurements quite useless.
- Everything else (and most importantly bathrooms) is done using custom ESP32-C3 devices that use SHT31 sensors to measure humidity and LD2412 for movement sensing. Also using the same adapter PCB for powering.
Thermostats are synchronized across all the devices with HA scripts. The HRV specifically uses its own wired temperature sensor to determine if it should enable heat recovery ("free cooling mode"), since its extracted air temp is always a bit lower than room temp (laziness :-)). "Current temperature" template fed to other heaters is derived from multiple room temperatures (currently using an average), with rooms "ignored" if AC is heating there (or was turned on recently). Ventilation has 2 modes set up - 20% and 80% - with the latter toggled by a bathroom humidity threshold.
There are 3 remaining things I want to set up:
- auto switching to gas heating if it's cheaper / the house is running on batteries - so far I've only imported electricity / gas prices into HA and quickly realized that I'm missing a power monitor on the heat pump circuit
- dampening of air ducts to reduce the temp drop when high humidity extraction boost gets triggered
- using more of the HRV range by auto-adjusting fan speed depending on real CO2 values - there's max 2 ppl at the house most of the time, so even at 20% the HRV is quite wasteful
This is so cool and something I’ve wanted for a long time, but it isn’t quite right yet (for what I personally want in an app like this.)
I am your target market, and I’d buy the lifetime /annual sub in a second if it had these features:
I want control of the SR sequence or, I want to know what SR algo you are using and know it is best practice model. The landing page says 4 sends, but that isn’t true SR.
The next thing, I’d want to see all of my “cards” or information pieces when I am logged in, so I can see and edit and delete and keep the database clean with a total view of content. The next thing I’d need (maybe you have this?) is for the email to effectively be a flash card, where the email content is the front and a link the email takes me to the “back” of the card so I can’t use cloze delete and other techniques. The last thing is bulk upload of content via a csv so I can bulk import mochi /anki / llm generated content.
I wish you luck with this and would (selfishly) encourage you to not ship so many different things, and instead encourage you to pick one and make it best in class for niche users like me who would spend and spend on premium solutions, but won’t spend on superficial implementations.
I’ve been working on a free language learning app that uses a local AI model to translate words and sentences. You can read sentences, short stories, or watch YouTube videos with translatable captions. It mostly follows the principles of the comprehensible input theory. Think of it like LingQ, but free and with any language combination.
The main goal is for the app to run entirely on the client side and stay completely free.
Personally, I can’t stand Anki or Duolingo—I’d rather read actual sentences that are fine-tuned to my level.
I can "imagine" what a building looks like, but I can't see it. Even our language makes this hard to describe, and the reason why I didn't realise that this isn't how other people do it, because most of our language for remember how something looks uses terms that implies we see them, so I just assumed until a few years ago (and I turn 50 this year) that that was just a metaphor to everyone else too.
How do you imagine (there's that word again...) blind people remember places they can't see?
To me, the notion that I'd need to "see" what the building looks like to draw it is bizarre because I know where everything is in relation to each other, and their shapes, so why would I need to see it?
EDIT: increasingly, and in part drive by split-brain experiments, I tend to think that a whole lot of what we see as our conscious decision-making, are instead shallow retroactive attempts by parts of the brain at rationalising decisions largely already taken autonomously into a cohesive self/ego.