HNHacker News
TopNewBestAskShowJobs

3uler

131 karma · joined June 16, 2017

submissionscomments
3uler··on I'm becoming AI-blind
Idk I’m starting to have have trouble with the frontier models and a couple how to write like a human skills…
3uler··on Cursor removed cost information from the usage page and CSV export
Calling per token usage of the US closed source labs has always been funny to me, we have a clear model of what it actually costs to host these models from open models.

Your subscription is not subsidised, it is just closer to the actual cost of the model…

3uler··on Google has abandoned Google News?
Half decade? Its been getting worse for as long as I’ve been conscious.
3uler··on Aurora DSQL: Scalable, Multi-Region OLTP
I feel like a lot of the trade-offs forced on you by dsql, you are forced into when scaling Postgres.

Like if I’m an SRE/DBA and you running massive update queries in a single transaction, that is the first thing I’m getting rid of.

3uler··on Our position on open-weights models
> industrial scale distillation attacks

You mean the industrial scale distillation attack you perform on the entire corpus of human knowledge… idk, call me cynical but you reap what you sow.

3uler··on Kimi-K3 on HuggingFace
American dates are sooo fucking annoying…
3uler··on Ask HN: How would you harden AI changes to a 1M-line legacy SaaS before review?
idk, I'm of the opinion that you own the code you wrote with AI.
3uler··on The new rules of context engineering for Claude 5 generation models
Markdown instructions are useful as durable context, but they do not replace programming. Their job is to keep intent out of a transient chat so both the developer and agent can revisit it. The actual constraints still belong in code: types, tests, linters, and dependency rules. Natural language can explain why a boundary matters; executable checks determine whether a change respects it. That feels more like a persistent layer around the programming workflow than a new programming language.
3uler··on The new rules of context engineering for Claude 5 generation models
I agree. I keep agent memory off. In my experience, it remembers the wrong details and applies them in the wrong places. Project context should live in the repo, where the whole team can find it, review it, and correct it.

It should be treated as an explicit artefact of the codebase for Humans and Agents to work with.

3uler··on The interesting part of an Agent Harness is what you add on top
I’ve never felt comfortable with how people use Claude Code or Codex as synonyms for Harness. It’s not wrong, it just reduces the harness to the least interesting part: the runtime. It’s the part you have the least control over. Most of us use what ever subscription makes most sense or whatever our company has settled on.

To me the interesting part of the harness is what you add on top, Brigitte Böckeler calls it the outer harness. I prefer the term Your Harness, your harness breaks down into two things: Skills and Loops.

Skills change what the agent knows and the loop changes how it works and checks itself. Both can mostly be encoded in Agent Skills

Your Harness should mostly agnostic to the agent you use.

3uler··on Stripe and Advent have made a joint offer to acquire PayPal – sources
In my experience, having worked at several large PSPs, it is Mastercard and Visa who are the morality police. Stripe are just enforcing the rules insisted on by them, and PayPal forgot about them 3 layoffs ago.
3uler··on How to stop Claude from saying load-bearing
Your tells, are just someone’s good writing now in the training set. It’ll be a moving target with each model.

I use the humanize skill to clean up AI written work before handing it over to colleagues.

https://github.com/blader/humanizer

3uler··on Fintech Engineering Handbook
For me quickly scanning over the article, the fact that floating points were even presented as a possibility was an immediate red flag. And I pretty much stopped taking the rest of it seriously.
3uler··on Salesforce to Acquire Fin (formerly Intercom) for $3.6BN
Sure but I would expect an exit of like 20b or just list and go public…
3uler··on Salesforce to Acquire Fin (formerly Intercom) for $3.6BN
What the hell? That is so cheap? I would value this at least as much as cursor? What gives?
3uler··on How to earn a billion dollars
I don’t know if this take is just naive or dishonest…

building something people love can make you a billionaire, but most billionaires did not build something people love, and most people who’ve built something people love are not billionaires.

3uler··on Doing nothing at work
This makes me never want to work on large engineering teams ever again.
3uler··on 32GB of DDR5 now costs $375 – AI shortage continues to squeeze PC building
Makes the Framework Desktop look like a bargain.
3uler··on DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost
https://github.com/anomalyco/opencode/pull/14743
3uler··on Migrating from Go to Rust
Golang is an amazing runtime with a bad language, one that conflates simple with easy. I view it the same way I view Java: a fine choice for a corporation, but nothing to love. Although Java’s gotten a lot better lately.
3uler··on DeepSeek reasonix, DeepSeek native coding agent with high caching and low cost
Opencode has really bad cache stability issues that they seem uninterested in fixing at the moment.
3uler··on How to convert between wealth and income tax
Also when most of you income comes from your wealth, your income tax rate is effectively 0%…

So complaining about having to contribute to the society that gave the conditions for your vast wealth is going to get you 0 sympathy

3uler··on Deno 2.8
But if you look at the node compliance tests, deno has better compliance now days…
3uler··on When Rails-way does not work anymore?
What do you mean? The whole point of Ruby on Rails is the rails way? Also the problems you are describing are not new and the community settled on adding some sort of service layer

https://shopify.engineering/shopify-monolith http://sporto.github.com/blog/2012/11/15/a-pattern-for-servi...

3uler··on If AI writes your code, why use Python?
I’ve always found Ruby to be way more readable, what keeps me using python is the depth of libraries is unmatched.

So unless you’re into burning tokens having AI generate untested libraries, I’d stick to using the most idiomatic tool for the problem you are tackling.

3uler··on Fake building: Claude wrote 3k lines instead of import pywikibot
Tbh that is some engineering teams I’ve worked on…
3uler··on DeepSeek v4
These models are open and there are tons of western providers offering it at comparable rates.
3uler··on Show HN: Daemons – we pivoted from building agents to cleaning up after them
I can not find a description of how it works on the site, magic hands daemons !

Cool story, but what runs when?

3uler··on High-Level Rust: Getting 80% of the Benefits with 20% of the Pain
Fix them if needed, the OP’s point is that for a lot of applications it is not needed.

For most cases you will still be comfortably in the JVM/golang performance window.

Rust is great language, fighting the borrow checker sucks, don’t do it if you don’t need to.

3uler··on Claude Sonnet 4.6
Do you not value your time? Paying a 100 bucks for a Claude max subscription is well worth it
Page 1 of 3Next →