HNHacker News
TopNewBestAskShowJobs

coffeecoders

899 karma · joined June 2, 2025

Full-stack engineer (North America, Pacific time) with deep production experience in RabbitMQ and Kafka at scale.

When not writing code, I'm either breadboarding electronics or hiking in the mountains.

Email: ntemail (at) protonmail.com

submissionscomments
coffeecoders··on AI World Clocks
Napkin math:

9 AIs × 43,200 minutes = 388,800 requests/month

388,800 requests × 200 tokens = 77,760,000 tokens/month ≈ 78M tokens

Cost varies from 10 cents to $1 per 1M tokens.

Using the mid-price, the cost is around $50/month.

---

Hopefully, the OP has this endpoint protected - https://clocks.brianmoore.com/api/clocks?time=11:19AM

coffeecoders··on Why is there no Uber for plumbing/HVAC? (and why there ought to be)
A lot of these "Uber for X" conversations miss a deeper point. The software only works as a force multiplier when the underlying labor is modular and fungible. Ridesharing works because every trip is basically the same API call i.e. person A to location B.

Plumbing or HVAC isn’t like that. Every house is a legacy system with zero documentation and different failure modes. Half the job is reverse engineering decades of homeowner hacks, and mystery piping. The other half is risk management.

If Uber had to deal with "each car is different, some are from 1978, half the wiring was DIY, and if you route one trip wrong the customer’s house floods," it would look very different.

The heterogeneity of the physical world is the limiting factor. Basically, devops for building (instrumentation, diagnostics and standardization).

coffeecoders··on You should write an agent
Yeah, it’s a never-ending curve.

I built an 8-bit computer on breadboards once, then went down the rabbit hole of flight training for a PPL. Every time I think I’m "done," the finish line moves a few miles further.

Guess we nerds are never happy.

coffeecoders··on Do not accept terms and conditions
Your site gives me ssl error.
coffeecoders··on Why a 'Boring' Life Might Be the Happiest One
This line is often used to downplay how much economic stress shapes mental health.
coffeecoders··on Why a 'Boring' Life Might Be the Happiest One
Sure, for most of human history, survival was a daily concern. It's just unnatural that in a time of record productivity and abundance, many still feel the same survival anxiety.

My point isn’t that we’re worse off than pre-industrial farmers. It is that progress was supposed to free us from that precarity, not repackage it with Wi-Fi and subscriptions.

coffeecoders··on Why a 'Boring' Life Might Be the Happiest One
The idea of slowing down and finding joy in small routines assumes you've already escaped the constant anxiety of bills. It’s hard to be at peace when your baseline reality is a few paychecks away from chaos.

There’s a quiet cruelty in how modern economies make stability feel like a luxury. Owning a home, having savings, or even time to be bored used to be normal milestones, now they are privileges. Until the basics are within reach again, the “boring” life will remain a fantasy for many who’d give anything for that kind of peace.

Most of us here work in tech and make six figures, but not everyone has that luxury.

coffeecoders··on M5 MacBook Pro
Wanted to see if it's worth upgrading from the M1 Pro, but I don't see the M5 Pro in the compare dropdown. https://www.apple.com/macbook-pro/compare/
coffeecoders··on Show HN: Timelinize – Privately organize your own data from everywhere, locally
Love the grind! One suggestion would be to add a demo link with some test data so we can see it in action.

I am also slowly "offlining" my life. Currently, it is a mix of synology, hard drives and all.

I have always thought about building a little dashboard to access everything really. Build a financial dashboard[1] and now onto photos.

[1] https://github.com/neberej/freemycash/

coffeecoders··on The AI bubble is 17 times the size of the dot-com frenzy and four times subprime
The troubling part is we are in the AI bubble whether we opted in or not.

The downstream effects are everywhere - real estate (data centers), energy (power consumption), education (automated tools), employment (tech layoffs), and equity markets (Nvidia propping up the spoos). It's not one sector this time, it’s systemic adjacency.

During the dot-com crash, retail got burned. During subprime, it was the global financial system. Today, AI exposure is distributed among mega-cap public companies, a16z-style VC funds, sovereign wealth, and shadowy circular debt deals between hyperscalers and AI startups.

Even if you are not betting on AI, your retirement fund is.

coffeecoders··on Jeff Bezos says AI is in a bubble but society will get 'gigantic' benefits
@dang

Title needs to be changed to something like

"Bezos says AI is in industrial bubble yet promises huge benefits"

coffeecoders··on Jeff Bezos says AI is in a bubble but society will get 'gigantic' benefits
Every bubble looks obvious in hindsight. The dot-com crash left behind Amazon and Google. The crypto crash left behind Coinbase and a few real revenue generating companies. If this is the AI bubble, then the survivors are going to look very obvious in a decade, we just don’t know which ones yet.
coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
My washer has this 'Eco' mode, that is supposed to optimize time, water, and energy usage based on load weight and how dirty my clothes are. It finishes anywhere between 30 to 50ish minutes. Same settings an d all.
coffeecoders··on Ask HN: What are you working on? (September 2025)
Such a good UI.
coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
Uber and Airbnb solved real pain points: global availability, predictable, nice interfaces, cashless payments, reviews etc. These are the things small, local systems struggled to provide consistently.

Also, my point isn't that the current system is objectively better. It's that scale and convenience created network effects that make it "essential" in practice, even if it’s buggy, slow, or worse in some respects.

coffeecoders··on Ask HN: What are you working on? (September 2025)
Nothing extraordinary like yall.

I've been down a prime numbers rabbit hole. Trying to see the largest prime I can generate in a browser.

coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
Convenience, consistency, and global. They solved real pain points, even if the cost and ethics are now questionable.

It’s like a software library that’s buggy and bloated but everyone depends on it because rewriting it from scratch is harder than dealing with its flaws.

coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
Yeah, the valuations are crazy, but think about it this way: the app economy and gig economy seemed absurdly speculative at first too. And now they’re worth trillions.

Early stage hype overshoots, sure, but sometimes it’s just pricing in things the old frameworks can’t even model yet.

coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
Totally get the frustration. Uber and Airbnb definitely have their flaws. But the sheer scale of users (200M last I checked) shows that a product can be both polarizing and essential at the same time.

Kind of like how we’re seeing with LLMs: not perfect, often overhyped, but undeniably useful at scale.

coffeecoders··on Instant Checkout and the Agentic Commerce Protocol
Hackernews used to dunk on Airbnb, Coinbase and Uber. Now they're part of our daily lives. Feels like we’re watching the same arc play out with LLMs.

I wonder if we’re seeing the same pattern repeat here with ChatGPT becoming big.

Outrage followed by inevitability.

coffeecoders··on Windows ML is generally available
Ollama started out as the clean "just run a model locally" story, but now it’s branching into paid web search, which muddies the open-source appeal. Windows ML goes the other direction: deep OS integration, but at the cost of tying your stack tightly to the Windows ecosystem, very reminiscent of DirectX.

Now, the real question is whether vLLM/ONNX or just running straight on CUDA/ROCm are the only alternatives or we are all trading one vendor lock-in with another.

coffeecoders··on Ollama Web Search
On a slightly related note-

I've been thinking about building a home-local "mini-Google" that indexes maybe 1,000 websites. In practice, I rarely need more than a handful of sites for my searches, so it seems like overkill to rely on full-scale search engines for my use case.

My rough idea for architecture:

- Crawler: A lightweight scraper that visits each site periodically.

- Indexer: Convert pages into text and create an inverted index for fast keyword search. Could use something like Whoosh.

- Storage: Store raw HTML and text locally, maybe compress older snapshots.

- Search Layer: Simple query parser to score results by relevance, maybe using TF-IDF or embeddings.

I would do periodic updates and build a small web UI to browse.

Anyone tried it or are there similar projects?

coffeecoders··on How to motivate yourself to do a thing you don't want to do
I find it interesting how a lot of this advice overlaps with the same tricks we use in software engineering to tackle big problems. Breaking things into smaller chunks or even gamifying with streaks is basically the human version of agile sprints.

Sleep, diet, and stress are like "system dependencies".

coffeecoders··on KDE launches its own distribution
Yes, the key difference is GNOME has strong downstream partners that treat it as the default (e.g. Fedora Workstation, Ubuntu). This way GNOME gets a lot of testing, polish, and feedback without having to maintain its own dist.
coffeecoders··on KDE launches its own distribution
Without being too negative, I'd like to point out that Neon, ElementaryOS etc tried the same thing. A project thinks we need our own distro but ends up pulling resources away from improving the desktop environment itself.

GNOME doesn’t maintain Ubuntu or Fedora, but it still dominates the Linux desktop experience.

coffeecoders··on Experimenting with Local LLMs on macOS
It is less about conversion and more about extending ANE support for transformer-style models or giving developers more control.

The issue is in targeting specific hardware blocks. When you convert with coremltools, Core ML takes over and doesn't provide fine-grained control - run on GPU, CPU or ANE. Also, ANE isn't really designed with transformers in mind, so most LLM inference defaults to GPU.

coffeecoders··on Experimenting with Local LLMs on macOS
I agree that it's kind of magical that you can download a ~10GB file and suddenly your laptop is running something that can summarize text, answer questions and even reason a bit.

The trick is balancing model size vs RAM: 12B–20B is about the upper limit for a 16GB machine without it choking.

What I find interesting is that these models don't actually hit Apple's Neural Engine, they run on the GPU via Metal. Core ML isn't great for custom runtimes and Apple hasn't given low-level developer access to the ANE afaik. And then there is memory bandwidth and dedicated SRAM issues. Hopefully Apple optimizes Core ML to map transformer workloads to the ANE.

coffeecoders··on I kissed comment culture goodbye
I feel the same way. Back in the day, I spent a lot of time on namepros and xda developer forums, and a few of the friendships I made there have lasted for years. I still talk to some of those folks weekly, and even meet a couple in person now and then.

It does feel harder to build that kind of connection today. Maybe it’s the anonymity, maybe it’s the sheer volume of noise, or just the way platforms are designed now. The sense of community that used to form around niche forums seems a lot rarer.

Maybe it was also my age. I was a teenager back then and more open to forming those connections.

coffeecoders··on 1TB Raspberry Pi SSD on sale now for $70
I think that’s the real distinction: the Pi isn’t interesting as a "cheap desktop", it’s interesting because of GPIO + the huge ecosystem of HATs, sensors, and weird little projects.

If you just want a small, cheap computer, x86 boxes are better in every way.

But if you want something you can hook wires to, fry once in a while, and find 20 blog posts about someone doing the same, Pi still has a niche that’s hard to replicate.

All I’m saying is that Pi projects can get really expensive if you want to do anything beyond a basic install like Pi-hole and leave it as-is. Even simpler projects - say, connecting a motor or a servo require additional components like a power module or distribution board, which quickly adds up.

coffeecoders··on 1TB Raspberry Pi SSD on sale now for $70
This is my recurring gripe with these "mini computers":

Power cable? Not included.

SSD adapter? Not included.

Needs a mini HDMI to HDMI cable.

Needs a cooling case (absolutely required for Pi 5).

It feels a bit like Tesla’s "gas savings" pricing deception. The Pi looks cheap on paper, but the actual usable cost is buried under "essential extras" [1].

[1] https://www.pishop.us/product/raspberry-pi-5-16gb/?src=raspb...

← PreviousPage 3 of 5Next →