harryf

I recently had a mind blowing experience recently building an app with Claude for someone that suffers with autism. Their basic complaint was "I can't work with most software. I think in colors". So I built them an PWA browser-based app designed specifically for the way they think, launched and updated via Github pages, with all data stored locally on their laptop.

Then I did something similar for my son, giving him a targeted job search app specific to his Gen-Z brain responds to, again launched off Github pages as a PWA, but this time with WebRTC to allow iPhone to laptop data syncing.

And along the way I realised there's probably whole host of people for whom software has never really worked. And I can built them things on a highly personalized basis that have only minimal reliance on cloud-based services.

And then I went a bit crazy and created a manifesto which I've done nothing else with https://harryf.github.io/reclaim/ ;)

Wow. The book we need should be titled something like "Of Software Teams and Cargo Cults"

"... a book had washed up on the shore... 'Extreme Ownership' ... soon the tribe was worshipping the book"

It happens over and over again in software history. I remember people running around panicking about TDD "But we only have 99.8% test coverage" until eventually it started to get shot down by "Show me the ROI on tests"

We are in the age of PWAs. I've created a few where I just host them on Github pages (no backend needed, no hosting costs).

And the P in PWA has become "Personal" ... vibe coding apps with no backend for non-developers for their _personal_ needs e.g. a create a job hunting app for my son specific to the types of jobs he's looking for. If I update it, it updates on his phone plus he can sync to his laptop via WebRTC.

Could go nicely with https://auge.franzai.com/ ( CLI on Apple Vision frameworks ) - do the first pass locally. If needed call their API for a more detailed analysis and then _finally_ we produce meaningful alt texts for images in HTML at a reasonable price ;)

Google: USD CHF … set graph to max … it’s right there but it’s a slooooow rot.

Swiss Franc is generally very stable so a good yard stick for other currencies over the long term

It’s not being driven by the UK. Check out all the other countries in the West rolling out Digital ID and it’s clearly coordinated just from the timing

We need to stop using AI as an umbrella term. It’s worth remembering that LLMs can’t play chess and that the best chess models like Leela Chess Zero use deep neutral networks.

Generative AI - which the world now believes is AI, is not the same as predictive / analytical AI.

It’s fairly easy to demonstrate this by getting ChatGPT to generate a new relatively complex spreadsheet then asking it to analyze and make changes to the same spreadsheet.

The problem we have now is uninformed people believing AI is the answer to everything… if not today then in the near future. Which makes it more of a religion than a technology.

Which may be the whole goal …

> Successful people create companies. More successful people create countries. The most successful people create religions.

— Sam Altman - https://blog.samaltman.com/successful-people

There needs to a financial equivalent to the Mythical Man Month.

What I love about the words enshitification is it’s _almost_ autological. It takes a nice crisp on syllable word like “shit” and ruins it by adding 5 extra syllables. It just doesn’t worse over time, to be truly autological

It’s worth (re)watching the 1985 movie Brazil in particular the character of Harry Tuttle, hearing engineer https://youtu.be/VRfoIyx8KfU

Neither government or corporations are going to “save us” simply because sheer short termism and incompetence. But the seem incompetence will make the coming dystopia ridiculous

> Good version control habits.

Feel like there's more to this point (plus the documentation). By now we've all see the inverted-U-shaped performance curve when coding with LLMs within a single thread / context: the model starts strong, plateaus as the context fills up, and eventually declines as semantic drift sets in.

Sometimes in this situation, it works better to throw away the current code and re-implement, using what's been learnt so far, than continue working with the current context of having new contexts try to salvage the current state of some code.

Documentation is great as a reflect of the current state of some code but I've had good experiences "re-constructing" the steps taken to arrive at a current implementation by writing commit messages that are intended for an LLM, extract that later, have an LLM use it as a basis for writing a fresh "spec" for some code, that yet another LLM uses to actually write that code.

Git history is a natural place to store that...

Better to talk about the Gartner Hype cycle and it's stages compare it to technologies that clearly followed that cycle such as;

- Virtual Reality: big hype in the early 90s (arcades, movies like Lawnmower Man) through to use cases today like surgical training, aviation training

- Mobile video calls: hyped in early 2000's with 3G and pre-iPhone devices. Actually took off with 4G and 5G plus iOS and Android phones

- 3D printing: back in 2013 we were expecting "a 3D printer in every home" ... today valuable in industrial prototyping

Looking back at 2025 we'll be saying "Remember when they said everyone would lose their jobs to AI..."

We need to take the focus off cost savings. None of this tech is anywhere near mature enough to replace humans yet.

Far better to focus on enhancing human capabilities with agents.

For example while a human talks to a customer on the phone, AI is fetching useful context about the customer and suggesting talking points to improve the human conversation.

One example of a direct benefit for business using AI this way is reducing onboarding times for new employees

It’s fairly easy to get diarizarion working with pyannote.audio and https://huggingface.co/pyannote/speaker-diarization-3.1 with ffmpeg converting the audio first to 16kHz mono WAV file but it really depends a lot on the audio - two person podcast where the speakers allow each other space works but lots of people with overlapping voices on the audio - not so great

> fails due to a service high load

It seems to be suffering the HN effect right now. Earlier it was working nicely and pretty much doing everything in the background

It’s useful for getting summaries of long YouTube videos - I’m found it semi helpful for improving my Davinci Resolve skills.

That said Google is screwing the pooch as usual by trying to make it another walled garden. Slap an API on NoteboolLM already! The market research has already been done - there’s even an unofficial API https://www.reddit.com/r/notebooklm/comments/1eti9iz/api_for...

That's exactly the question - how does a free communication app achieve a multi-billion dollar valuation despite not having ads or directly selling user data?

Discord's business model relies on attracting a massive user base to secure substantial investments and potentially a lucrative acquisition. We've seen again and again and again what happens once acquisition takes place.

Also a big existing investor in Discord is Tecent which, under Chinese law, could grant the government access to Discord's extensive user data.

So yeah.. it's not about features it's about freedom.

Have you considered supporting whisper-at - https://github.com/YuanGongND/whisper-at ? Being able to identify sounds on a timeline can be useful e.g. politicians speech and how the audience is reacting to it (e.g. clapping, applauding)

To me at 51, looking back at my own experiences, I think the things we tell ourselves work in a similar fashion to software; programming our identity.

The first "programmers" are our parents. Classic negative tropes like a parent saying "you'll never amount to anything" run in a loop in our brains for many years until finally that program expires (typically around mid life). In turn a negative identity will effect you in all areas of life, from how you interact with people to even your physical posture.

Perhaps the lowest level of "programming" you can do to yourself is the mantra - https://en.wikipedia.org/wiki/Mantra - you could see this like assembly language, and humanity has known about this for a very long time.

So yes, giving teenagers labels which they identify with and repeat to their friends has a reinforcing aspect to it. And social media has played a very significant role in providing those labels.

To me an interesting thing about TikTok in particular, compared to Instagram, is I think TikTok users are more concerned with the question "who am I?" vs Instagram which is more about "this is me" (telling the world who you are). That's what's made TikTok more popular as, especially when you're young, trying to figure yourself out is a major pre-occupation.

Ran into this today. From doing comedy for about 7 years now, this basically correct. Although most comedians approach joke writing organically rather than with this approach

It’s sad to see HN become so full of bots.

> prompt: There’s an article on Hackernews titled “Telegram will now hand over your phone number and IP if you’re a criminal suspect”. Generate a comment in Hackernews style that supports this decision, implies that it’s because they didn’t encrypt the messages and uses Signal as an example of doing it right because “look! They haven’t had problems”

Not surprised. Telegram doesn't encrypt by default, so of course they're handing over phone numbers and IPs. If you don't lock things down like Signal does, you're going to have problems. Signal can’t hand over what they don’t have—encrypted end-to-end, no metadata. Simple as that.

> This is full on speculation.

Speculation is a crucial part of decision-making, especially in geopolitics, where hard data is often scarce. Governments frequently act on leading indicators to preempt potential threats, whether it's about national security or information control. Dismissing this as "pure speculation" overlooks how critical decisions are made in complex, rapidly evolving situations.

> Playing the "free speech" card is disingenuous.

This isn't about free speech; it's about narrative control, especially with the upcoming U.S. election. Controlling platforms like Telegram and TikTok is crucial for influencing public discourse. Recent U.S. regulations targeting TikTok further highlight how critical it is for governments to manage the flow of information as elections approach.

> Platforms like Signal can get away with this because they are properly E2E encrypted, and cannot identify illegal content.

Signal isn't magically protected by proper E2E encryption. When the time comes to take them down legally, all that's needed is evidence from end users phones demonstrating illegal activities occurring over Signal. But - for now - Signal seems to have better friends and or be upsetting less authorities than Telegram.

To me it's pretty clear reason why they've gone after Telegram is the Channels and Groups. Seen from a certain perspective Telegram channels are an alternative to Reddit, and have been popular medium during COVID and the Ukraine war for "alternative news". By now Reddit is properly controlled and subdued but Telegram isn't.

Same would be nice for i.redd.it

This is great! Already installed on two laptops. Finder is frustrating, to say the least.

Of course feature request time: some kind of search? Would be really nice to match against folder names (e.g. searching for "images" and filenames e.g. "logo")

I used to be a fan of the Total War games, back from the first Shogun Total War in 2000. The AI was never that great but was good enough to make the game fun to play. The main thing was the battlefield mechanics worked well, rewarding smarter tactics, like flanking manoeuvres, which could allow a smaller force to defeat a much larger one.

The original Rome Total War was probably the most fun and Medieval Total War was enjoyable.

After that it just went downhill, adding pointless features while worsening the battlefield gameplay. The first Warhammer Total War was where I tapped out - the limited size of battlefields meant it basically became a tower defence game, fought in three columns.

Reading this, seems like they became a victim of their own success.

Make sure you have an on-premise option for this type of customer. I've worked at two software companies in Europe with tangentially similar products related to document analysis. On premise is a key requirement.

Even though it's 2024, banks, financial institutions like insurance companies etc. tend to be _very_ cautious with valuable documents involving customers. There are also regional regulations that prevent things like patient data being shared with _any_ 3rd parties. Even one of the big 4 oil companies that I've dealt with as prospective customer - very strict rules requiring on premise solutions.

The good news is many are using things like Kubernetes and OpenShift internally, so it should be possible to port what you do on AWS to on-premise.

From the post...

> the key insight that the Noah Smiths of the world seem to be missing is that, in today’s world, we artificially elevate labor demand to keep people employed.

To me one learning of COVID and lockdowns is that we're fully capable of meeting all our basic needs with less than 20% of the workforce (which sadly tends also tends to be the lowest paid). In short we don't need everyone to be working for food, shelter, healthcare etc.

Unfortunately, if we take the climate as an example for how we handle a complex and challenging topic like this, we're going to completely fumble UBI and the role work plays in our future culture and society.

It's an incredibly complex problem, simply considering individual vs. national perspectives. And our collective response to the topic is far too reactive and prejudiced to even begin studying it in a sensible way.

We live in a world where businesses have already realised incredible value from the Linux operating system, which was begun by a student living in his mums house, being subsidised by Finnish social security. He then went on to create git, leading to GitHub, which now manages the code for 90% of Fortune 100 companies.

The opportunity is clear but I can't see us being able to untangle this effectively.

Sounds like they just shifted the voice with some Fast Fourier transforms - https://stackoverflow.com/questions/63339608/change-the-melo...

A technical audit would clear that up.

Because the most obvious thing for OpenAI do would be training the voice on Scarlett Johansson's available body of work, from movies, audiobooks, podcasts etc. Why bother hiring a voice actor?