Every two months we seems to have a new SOTA now. What do you expect AI coding to look like four years from now?

Will it be like Go after AlphaGo, where even Ke Jie could only say "they no longer knew what they were fighting against"? Or will it just be like today, just stronger, with humans still driving the work and AI as a powerful assistant?

Comments URL: https://news.ycombinator.com/item?id=48624855

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48624622

Points: 2

# Comments: 0

As we'd already had plenty of reasons to move off of GitHub (downtime, a website that has gotten consistently slower due to massive increases in client-side JS without new features over the past decade, PRs that won't load once they get past 50 comments, contributors getting banned (without crypt-mining) leading to potentially-useful PRs getting black-holed, slow support, etc, etc), this is more of a warning for others than any kind of attempt to get help.

Of course Github is struggling these days with an influx in AI Agent accounts driving a huge increase in spam and other garbage, so I sympathize a lot with the folks over there. But none of that means we have to use the (historically excellent) free product they're offering, we can also...not.

For those who weren't aware, codeberg/self-hosted forgejo can import entire github repos including historical issues and PRs, comments, etc.

Comments URL: https://news.ycombinator.com/item?id=48624574

Points: 7

# Comments: 1

My workflow hasn't changed, same level of concurrency as usual, did they change their limits?

Comments URL: https://news.ycombinator.com/item?id=48624168

Points: 6

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=48623750

Points: 4

# Comments: 4

Most products are decent at creating first contact. Events, meetups, dating apps, friend apps, group chats, alumni networks, Slack communities, etc.

The part that fails is usually after the first meeting.

Someone has to follow up. Someone has to risk seeming too eager. Someone has to pick a plan. Someone has to coordinate calendars. If nobody does that within a week or two, the connection fades and everyone starts over.

I have been calling this "the second hangout problem."

The product idea is an AI go-between that helps people in one city turn one-off encounters into repeat plans and small groups. It would ask who you would actually want to see again after a meetup or small gathering, check for mutual interest, suggest low-pressure next plans, and help coordinate until the group has enough momentum to organize itself.

I am starting with the Bay Area because density matters. The product is not useful if people cannot realistically meet again.

Questions I am trying to think through:

- Is follow-through actually the bottleneck, or am I underrating discovery? - Would you trust an AI mediator for this if it stayed in a logistical/support role? - Should a product like this limit introductions on purpose to avoid swipe-style disposability? - What would make this feel useful instead of creepy?

Curious how people here have seen adult friendships actually form after the first meeting.

Comments URL: https://news.ycombinator.com/item?id=48622679

Points: 5

# Comments: 4

Let's look at how we got here, and the wall the industry is currently hitting.

- The Dream Phase (2010-2016). By the dawn of the last decade (2011), an interesting thing was happening. The two platforms, Wikipedia and Stack Overflow, had started gaining tremendous traction, folks were collaborating on these platforms to openly exchange knowledge. Looking back, this feels like a more ideal, community-driven path for humanity — one we abandoned for the centralized architecture we have today.

- The Disruption Phase (2016-2021). A perfect storm of unrelated events paved the way for AI. By 2017, new programmers were growing deeply frustrated by Stack Overflow's rigid policies, subjective question rejections, and senior coder pedantry. In retrospect, those strict moderators carved the first stones of what would later become Copilot and ChatGPT. If the community won't answer a beginner's question without downvoting it, a private LLM gladly will.

Add to this Google's landmark 2017 paper "Attention Is All You Need" which unlocked the Transformer architecture, and the forced isolation of COVID-19 in 2020. The ground was suddenly fertile for virtual assistants that could act as isolated developers' programming partners.

- The Hook Phase (2023-2025). The launch of ChatGPT left no doubt about how easy the "hook" would be. For non-technical folks, it was pure magic. It didn't take long for specialized LLMs like Copilot, Claude and Deepseek to become an indispensable part of the programmer's toolbox. Meanwhile, OpenAI was still advertising its "non-profit" roots, and the consensus was that this was purely about empowering humanity.

- The Endgame Phase (2025-present/future). AI companies had miscalculated a lot of things by this time. They were optimizing for the "long-term" but as John Maynard Keynes rightly said many years ago, "In the long-term, we are all dead". The VCs are losing patience today because while the technology itself has gained massive ubiquity and appreciation, the revenues aren't coming as fast. The hook had sort of worked but failed to fully work.

Most frontier models like Sonnet, Opus and GPT 5.5 are still running on 'subsidized mode'. The amount of monthly subscription they charge users (USD 10/20/30 per month) is a pittance compared to all the compute and RAM needed to run those "thinking..." and "pondering..." tokens. In order to truly show profits in the books and come out of subsidized mode, they must charge on the scaling of input/output tokens and that appears to be difficult. Very few companies might be able to sustain such unlimited budget for unpredictable hardware scaling, the recent Uber story shows exactly what happens when they try doing this.

The frontier models are trying to replace something which could never be successfully delegated or automated in entire human history - the highest cognitive skills of human brain like reasoning, deduction and logic. Yet, the efforts are on and the goals are long term. The conundrum is that if they stop subsidizing, the hook phase may be undone - there is a strong possibility of folks reverting back to older ways of Wikipedia/Stack Overflow or pivot entirely to open source dry/academic models like Llama and Qwen which can run locally on their own hardware. And yet, they also can't keep subsidizing and draining the funds indefinitely.

What happens when the subsidy mirror cracks?

Comments URL: https://news.ycombinator.com/item?id=48622280

Points: 10

# Comments: 1

What I notice is that LibreOffice can create very complex PDFs when the Microsoft Word document has:

1. tables

2. multiple columns

3. strikethrough text

I am thinking we should go straight from Microsoft Word to Markdown.

What is the right software for that?

Comments URL: https://news.ycombinator.com/item?id=48622123

Points: 4

# Comments: 6

Points: 4

# Comments: 3

I started coding, nah, writing markup and stylesheets when I was 10 or something with NetBeans which was weird but it was the IDE that the tutorial guy used, hoping to learn how to make a website, make money then buy a good computer so I can run a proper 3D game engine (instead of construct which I didn't like) and make games.

While at first I just wanted to make money, It felt good. It went like this for a year and I didn't make any money. And I think I stopped programming or more precisely did much less programming than before for some reason I don't remember. After a year I started again, by then I had learned that I don't have to use NetBeans and I can use notepad++ which was smoother on my 1GB-ram-mobile-chipset-broken-keyboard-black-monitor-tochpad-burnt-connected-to-TV-via-HDMI Acer laptop. And yeah my neck hurt looking at the TV from close distance.

I switched to English learning sources, primarily w3schools. And It went on like this, I learned some JS. I bought a course with info I already knew, but it filled some holes on basic things, for example I didn't know what a title tag is. I completed the freecodecamp.org courses even though I knew most of the stuff, until the JS part. I didn't know what functional programming, OOP, and mutation mean. I was told to not change a variable, I didn't, I passed. I didn't understand. I was hoping to get a job with getting a freecodecamp certificate. Then I went to get the React certificate, I hadn't worked with React before or any other framework. I learnt why frameworks are cool.

I participated in some middle-school competition, I chose the software topic for some reason and made a timer with C# and it got rejected. Next year same story with Game topic, I don't know if I even made anything. Then next year I made a question generator with js and it was rejected.

I got accepted to "shining talents" school, and people got surprised because unlike others I studied while I programmed and watched movies and played games, even in finals. People were very smarter than me and I found out I was a big frog in a small pond. But many of them didn't program. So I was mostly studying, but then after the first term exams, there was a golden time, when I only programmed and studied, mostly programmed and didn't play any games. Not because I forced myself to, because programming was much much more enjoyable than games to me.

That time passed, and not only I programmed less than the peak, I programmed less than ever. By the way, for 2 years, I would do 6 months of web dev, get bored then switch to game dev. And I participated in GMTK game jam every year. I made no interesting projects. Not even boring completed projects. Of course I completed a game once a year at the jam and that was it.

There was more studying, and then AI hype. Before AI hype I didn't like the way things went, I had to search simple things simple as loop syntax in JS (I kept forgetting syntax). Also I was only learning frameworks and making nothing. I was learning svelte at the time. After AI hype, I wrote literally no code. Sometimes I opened a HTML to program without AI, then I would just set up the basic stuff like fonts etc and then closing it. Even with AI I made nothing.

My friend told me to participate in national AI Olympiad. I got a national gold medal. And it was good because I felt even though my programming journey wasn't perfect, it has helped me, as I was saved by my practical score and became 1st practical. But I don't know what to do from now on. I don't seem to like AI. Even not sure if I like game dev anymore. Please help.

Comments URL: https://news.ycombinator.com/item?id=48621795

Points: 3

# Comments: 1

They look like Video Games sometimes too.

This reminds me of that case of a scanner used in an architecture studio that was changing numbers for others, and it turns out they were using content aware compression and decompression.

Maybe the commercial pressure to use AI caused them to apply an AI filter to upscale the images and they end up looking like videogames?

To be precise I'm looking at the Belgium - Iran game, it's a video feed that is clearly broadcast to all of southamerica. Don't know if that can help someone find it.

Comments URL: https://news.ycombinator.com/item?id=48621628

Points: 3

# Comments: 2

He took me to a quarry to fire Estes-style rocket cars, and all that.. he also managed to steal a Milicja (military police) siren from work, and put it on the back of my banana style bike. As an EE, he made it actually function.

Thanks to all the hackers who made us who we are.

Comments URL: https://news.ycombinator.com/item?id=48620502

Points: 286

# Comments: 44

Closest parallel I could think of is gift cards, which too do not expire in many places.

Comments URL: https://news.ycombinator.com/item?id=48619635

Points: 5

# Comments: 5

Comments URL: https://news.ycombinator.com/item?id=48619320

Points: 3

# Comments: 1

I am building a new product and need to have a robust queue system but I'm not sure what product to choose.

I've worked in Amazon before and usually AWS tools are the default go-to but in my own time for small project I've used things like: https://github.com/hibiken/asynq for basic tasks.

My concern with the above is that it's still in "early" development; from their README:

``` Status: The library relatively stable and is currently undergoing moderate development with less frequent breaking API changes.

Important Note: Current major version is zero (v0.x.x) to accommodate rapid development and fast iteration while getting early feedback from users (feedback on APIs are appreciated!). The public API could change without a major version update before v1.0.0 release. ```

I'm building something right now that will require something very robust; scheduling, long running workflows with several external dependencies (which may have failure and need retries and notifications, webhooks, etc)

Ideally I want to use the same tool for everything; so that would also include basic things like fan-out (like user singed-up: fan out to email service to queue welcome sequence, notify another service, make api request to another dependency, etc.)

There are so many too choose from: NATS, RabbitMQ, Asynq, AWS tools, Google Pub/Sub, etc.

There are so many options and I'm always very excited to explore tools but I need to pick one and us that across entire project for everything; since it won't be just me and having one interface/tool to learn reduces cognitive load.

What do you use normally and why?

(I use Go - if it matters)

Comments URL: https://news.ycombinator.com/item?id=48619311

Points: 2

# Comments: 2

With my current job and child care stuff I have kinda fallen out of contact with a lot of people I used to work with( I don't really see many people in person for meaningful conversations about professional things) so I'm also a bit isolated in that regard.

Comments URL: https://news.ycombinator.com/item?id=48619097

Points: 2

# Comments: 0

Points: 5

# Comments: 4

Comments URL: https://news.ycombinator.com/item?id=48618933

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=48618362

Points: 3

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=48618170

Points: 5

# Comments: 2