HNHacker News
TopNewBestAskShowJobs

pvtmert

571 karma · joined July 10, 2016

Tinkerer, hacker, researcher, & cyclist...

  DevOops Engineer by trade
  BSc. Software Engineer by training 
https://mert.akeng.in https://pvtmert.at.hn

https://meet.hn/city/49.6202283,6.1729364/

submissionscomments
pvtmert··on I ditched Docker for Podman
Docker is useful and running in production. (well, it was the only, before containerd got separated entirely and it was usable directly from K8s about when it was 1.16)

What you say is absolutely correct. If Docker keeps creating compatibility layers for it's competitors, it makes everyone to switch to a competitor. In this case, the competitor is Kubernetes as it's running in production for much larger scale (enterprise workloads) compared to Podman et. al.

Hence, it's the job of Podman, Kubernetes, et. al. to write their compatibility layer to provide a value-add for their customers.

pvtmert··on I ditched Docker for Podman
This is a great summary, thank you!
pvtmert··on I ditched Docker for Podman
Not podman user (but currently trying to install to give it a shot), this comment stream shows how even the documentation "randomly disappearing" on a project that claims in production-ready or stable state. (Or lack thereof)

On the contrary, docker documentation *is* stable, I had bookmarks from 10-years ago on the *latest* editions, that still work today. The final link may have changed, but at least, there is a redirect (or a text showing that has been moved) instead of plain 404/not-found.

This is a crucial part of the quality applications offer. There might've been 100s of podmans probably since Docker was launched more than 10 years ago, but none came close to maintain high-quality of documentation and user-interface (ie. cli commands, switches). Especially in the backward-compatible way.

pvtmert··on I ditched Docker for Podman
Why would docker should create such tool in the first place? It's the job of the target/destination to provide compatibility layer. In this case, Kubernetes already does with kompose.io tool.

kompose: https://kubernetes.io/docs/tasks/configure-pod-container/tra...

Also, technically docker-compose was the first orchestration tool compared to Kubernetes. Expecting former to provide a translation layer for the latter is rather unorthodox. It is usually the latter tool provides certain compatibility features for former tools...

pvtmert··on I ditched Docker for Podman
Both (K8s and Compose) are well defined schemas, hence the conversation is mere mapping via search & replace. Bunch of `sed` statements could do that, LLM is an overkill for the job.

Meanwhile, kompose.io exists, which is exactly what it does (but with Go templates as far as I can tell)

pvtmert··on I ditched Docker for Podman
I mean I get the overall benefits of having rootless and all, but the premise of security in this article is a bit misleading.

Obviously having a daemon running as root is larger attack surface than a program running as the user.

Going to the github.com/containers/podman/releases, the latest release is actually addressing a security risk that involves overwriting files of the host.

    # v5.6.1 (Latest)
    ## Security
    - This release addresses CVE-2025-9566, where Kubernetes YAML run by podman play kube containing ConfigMap and Secret volumes can use crafted symlinks to overwrite content on the host.
As always, the most secure computer is the one that is unplugged & turned off.
pvtmert··on Parallel AI agents are a game changer
Because most people have the context-window of 10 tokens, they do not read further than the first sentence (or two).
pvtmert··on Parallel AI agents are a game changer
> This is just software engineering?!?

Indeed yes. Although most places shipping software in a "software development" and/or "programming" fashion for many years.

Many, many places certainly do not do the engineering part, even though resulting product is a software.

pvtmert··on Parallel AI agents are a game changer
Is it me or the post sounding (showing!) that they haven't tried the mentioned approach in real life.

Because in real life, one agent tries to fix build issue with rm -rf node_modules and the other is already running a server (ie: npm server), conflicting with each other nearly all the time!. (even if it's not a destructive action, the second npm server will most likely to fail due to port-allocation conflicts!)

Meanwhile, what I found helpful is that: 1. Clone the same repo twice or three times 2. In each terminal or whatever, `cd` into it 3. Create a branch, run your ~commands~ prompts (each is their own session with their own repo) 4. commit/push then merge/rebase (also resolve conflicts if needed, use LLM again if you want)

Any other way multiple agents work in harmony in a single repo (filesystem/directory) at the same time is a pipe-dream with the current status of the MCP and agents.

Let alone being aware of each other (agents), they don't even have a proper locking mechanism. As soon as you make a new change, most of the editing (search/replace) functionality in the MCPs fail miserably. Then they re-read the entire file, just creating context-rot or over-filling with already-existing stuff. Soon you run out of tokens (or just pay extra for no reason)

> edit: comments mentioned that each agent runs in a VM isolated from others, kinda makes sense but still, there will be massive merge-conflicts unless each agent runs in a completely different set of service/code-base (ie frontend vs backend or couple of micro-services)

pvtmert··on Next.js 15 Markdown Boilerplate
Why not using marked.js or equivalent, plugging it with a Fetch API instead?

example:

  <html>
  <head><title>hello</title></head>
  <body>
  <main id=root ></main>
  <script src=./marked.min.js async defer ></script>
  <script>
  async function load() {
    let element = document.querySelector("main#root");
    await fetch("./index.md") 
      .then(resp => resp.text())
      .then(text => marked(text, { ... }))
      .then(html => {
        element.innerHTML = html;
      });
  }
  window.onload = load;
  </script>
  </body>
  </html>
since it directly injects the resulting html, you may set "unsafe HTML" option in marked (or markdown.js) and include <title> tags within your markdown document. this will _also_ rewrite the browser title(s)

> edit: formatting

pvtmert··on Next.js 15 Markdown Boilerplate
This is exactly how my (personal) page(s) work! I wrote it pre-covid.

Initially I was using the hash as you suggested, but the redirect/404 was flaky when I use github-pages as the host, hence I reverted to the query-string (? or search) as it creates a complete refresh. Which allows me to set up "virtual" routes like `/blog`.

Topping with worker/service worker, it is quite robust!

Link: https://mert.akeng.in/

pvtmert··on Amazon has mostly sat out the AI talent war
I usually do not agree with the Amazon leadership (well, recently they haven't been "Right A lot"!)

But I agree with the following statement Matt Garman gave recently;

    Amazon Web Services CEO Matt Garman said that using AI tools in place of junior employees was "one of the dumbest things I've ever heard" because these employees are "the least expensive" and "the most leaned into your AI tools."
It's because AI usually creates slop, without review these "slop" build up. We don't have infinite context window to solve the slop anyway. (even if we do, the context-rot has been confirmed)

Also, on average, Indian non-Tech employees who manages thousands of spreadsheets or manually manages your in-store cameras are much more cheaper than the "tokens" and the NVIDIA GPUs you can throw at the problem, at least for now and a foreseeable future.

pvtmert··on Amazon has mostly sat out the AI talent war
Ah, I just saw this, although I wrote a (much comprehensive) reply here: https://news.ycombinator.com/item?id=45097345

The full payment that requires pro-rates is even worse. They expect you to pay it fully back. (ie. with the deducted taxes included!)

I bet it is possible to profit from a such scheme if Amazon is able to declare that as a reversed-transaction (similar to VAT-refunds) at the end of the fiscal year.

pvtmert··on Amazon has mostly sat out the AI talent war
I joined in 2022 from a different location, there were 2 kinds of comp in terms of bonues, each split into 2 other;

1. Relocation package a. Lump-sum (7k EUR): You get certain amount of money, and you deal with your own move yourself. (Albeit with some reimbursement possible for the initial trips) b. "Other" (I don't remember the name): More supportive option, good if you have family & furniture to move. They essentially pay everything for you. c. Important: The 7k EUR was subject to the tax, hence I got taxed at 55% (EU) due to having no tax residency at the moment (obviously). Nobody ever mentions this. But the re-payment is with the tax-included, ie. you are expected to pay 7k back! 2. Sign-on bonus: This splits into 2-year period a. 1st year: 50% of the total bonus, transferred to your bank account on your first work day. b. 2nd year: Each month, you get 1/12 of the remaining 50%, essentially something like ~4.18% each month on the second year. c. The 50%/50% ratio may depend on the team/role/location, I heard some of the L4s joined to the team got split of 40%/60% (ie less in the first year) for reasons unbeknownst to me.

Conditions are pretty simple, if you leave (for any reason), you must repay monthly-pro-rated amount that you haven't worked given the total period is 24-months. ie. In Luxembourg, probation is 6-months. (Until) at the end of the probation, Amazon can just fire you for no reason. In this case, since the 2nd year sign-on hasn't vested yet, nothing to pay from that, but you must pay 1/4th of your "relocation expenses" and full half of (ie untaxed full amount divided by 2) sign-on bonus you receive on your first day. (ie. 25% of the total sign-on bonus)

Firstly, I know someone (a Greek national) who left Amazon during his 12th Month. Amazon demanded total of 4k+ euros from the guy, citing he hasn't finished his 12th month, hence the first half of his relocation bonus plus the 1-month of pro-rated sign-on bonus, before tax. As far as I know, it was more or less equivalent to his monthly gross salary, and he paid in installments.

Secondly, I heard someone joined from non-EU country in 2023, and essentially got laid off. But because she was in probation and obviously worker rights are much stricter in EU, Amazon just declared her as a probation-failed case instead of layoff. (She also got laid off within last 2 weeks of her 6-months long probation). Since she only got the residence permit recently, not having more than a few months (when unemployed as a 3rd-country national), plus Amazon demanded money to be paid back. As far as I know she contacted an labour lawyer and they basically advised her to go back and not to pay anything back as it becomes an international matter. And the costs/fees for such is much higher than what would Amazon get it back, hence she did what was suggested. Although it obviously burns the bridges but in this case, Amazon started the fire first...

---

As a result, the practices applied here falls no short of what you can hear from the news. As the company has no heart or soul, people are just numbers in a balance-sheet...

pvtmert··on Thoughts on (Amazonian) leadership
As a ~4 year tenured L5, I agree. There is annual performance review system called Forte. During forte, each person is rated in each leadership principles (LPs) by their peers and manager, possibly from away-team members.

As the parent mentioned, nowadays the scope is extremely limited. Citing Ownership (ie. this/that team has the ownership...) reasons. I see LPs are currently weaponised to limit promotions/pay-raises.

I am not a low-performer or anything, in fact, I proactively find problems and fix them. I do not like to complain (which most people do) and if it takes <=10% of my weekly working time (ie. 4 hours or less) then I create a ticket and add this to my personal backlog. Then, when I go through my backlog, I prioritize these things according to their predicted complexity and impact. I take the low-complexity/big-impact things and do those in a 4-hour period.

When I fix the things, I update the ticket and send a CR (pull-request) to the owning team of the package. I even have a script to pick out a recent committer in the repository to add them as an optional reviewer, which helps quite a lot, as most of the teams do not track their CR (pull-request) queues at all. (neither proactively or even with notifications)

Nevertheless, 3 years back to back, I have been rated negatively for the following LPs - Think-Big - Ownership - Bias-for-Action

Moreover, I have not only participated in Hackathons but even organised org-wide (Director/L8) hackathons & events. Not to mention that my manager, my skip, and their manager (director) all in different regions than mine. I didn't had any TPM in my org in my region either.

It super frustrating as an engineer as certain minor occurrences weaponised against (ie. an escalation from away team that I did not respond in time during my PTO, c'mon!) meanwhile I have 10 or more valid scenarios which keep being overlooked.

Which is why the quality is dropping. Because as an engineer, I do not see any return on my investment (time & effort). There are thousands of engineers like me, I see more and more people are silent quitting (rest & vest), elevating minor things as if they were grand problems, increasing the bureaucracy as much as they can. As the layoffs already showed to all, there is no job guarantee at all!

pvtmert··on Decoupling Agents from IDEs – Agent Client Protocol (ACP)
The Agent Client Protocol standardizes communication between code editors (IDEs, text-editors, etc.) and coding agents (programs that use generative AI to autonomously modify code).
pvtmert··on How to slow down a program and why it can be useful
macOS has embedded feature that allows slowing down network calls. Similar to the IP/netfilter suite Linux has. PF handles most of the magic behind the scenes.

Even better because there is an UI to configure it. You just need to download Xcode additional tools package, there is a NetworkConditioner.prefpane inside. Install that and various settings will show up in the regular System Settings/Preferences...

pvtmert··on A failure of security systems at PayPal is causing concern for German banks
I default to the Reader mode (Safari/Firefox) such a blessing to stop any/all videos, sounds, ads, and popups.

Putting the "text" back into HTTP, hyper text transfer protocol.

pvtmert··on A small change to improve browsers for keyboard navigation
Honestly, most sites these days built in such incompetent way I cannot imagine how much disabled users suffer...

Since there is no tariff adding <div> in react code, even my teammates are using these as pseudo buttons. As a user, I cannot even tell if it's a button or not. The cursor usually doesn't change or becomes "I-beam" (text selection). Only way to understand is click random places and wait for mysterious stuff to happen.

pvtmert··on Scientists unlock secret to thick, stable beer foams
maybe I am the unpopular opinion but I try to make it less foamy...
pvtmert··on Google will allow only apps from verified developers to be installed on Android
Meanwhile, I suppose a big "rollback" will needed in EU for the DMA (Digital Markets Act)
pvtmert··on Recreationally overengineering my location history
Even though it is at overengineering.dev, loved this pragmatic and simple approach:

    I have implemented the most advanced client-side error-handling technique ever: window.location.reload(). 
In larger (stateful) apps, this is probably not what you'd want. But works for small, stateless apps...
pvtmert··on Hacker and physicist – a tale of "common sense"
I am gonna be that guy but;

    A developer can slap a @RestController annotation on a Java class and magically have an HTTPS endpoint without knowing anything about TLS handshakes or certificate chains.
Is a bad thing. Most people only skims the surface of what they use in daily life. Obviously there are so many things that it is hard to do otherwise. But this job is one time only...

Which is why AI seemingly replacing junior engineers, because AI does the same thing, faster and cheaper...

Meanwhile, I as an engineer, see less and less appreciation and importance in my organization about "knowing" these stuff.

pvtmert··on Static sites with Python, uv, Caddy, and Docker
I see the Docker stuff here but Kubernetes is missing. Lacking required complexity for a simple _static_ personal blog.

Also while using Kubernetes, please use event-driven PubSub mechanisms to serve files for added state-of-the-art points.

/pun

pvtmert··on Static sites with Python, uv, Caddy, and Docker
delayed
pvtmert··on Fmllm: 4mb training data, 100mb model, Fibonacci embeddings, near-coherent. WTF?
it certainly seems interesting, however with 4mb of text, i believe/think you can also create a _legible_ enough markov-chain that will output more/less similar sentences, minus punctuation of course.
pvtmert··on Vibe coding creates a bus factor of zero
Yes, the engineer who coded is already dead. No bus necessary.
pvtmert··on Universal Tool Calling Protocol (UTCP)
Soon, they will discover APIs. (Application Programming Interfaces). Where programs interact with each other with given contracts/protocols/prototypes.

/pun

pvtmert··on Sequoia backs Zed
On the bottom right, there is a button looks like magic emoji (). Click that and it will show general AI settings. You can choose _Eager-Completion_ rather than trigger-based one.

As a user, most IDEs/Editors currently show in _eager_ way. For example, VSCode by default shows ghost-text as well as Amazon Q extension too. Usually disabling those disables the AI completion completely.

Meanwhile I like Zed's approach that you can trigger completions with Alt+Space, not burning through your "tokens" in free-tiers. They also provide a free-tier completions, as well as _next-edit suggestions_.

*) Next-edit suggestions: When you edit some piece of code repeatedly, it suggest to do similar on the next few instances, with context awareness, quite nice feature saving several keystrokes every single time.

pvtmert··on Sequoia backs Zed
although I tried the collaborative features I unfortunately cannot say that I used these extensively for now. Although I found them quite nice, well integrated, seamless, and straightforward!

The issue is again people, they don't wanna change their _archaic_ workflow, stuck with inefficient -copy/paste- loop to the chat (ie. Slack) and back.

The story in the article went a bit too far that I agree, but I guess that is their north-star vision. Current implementation allows you to "join" a workspace session shared by someone, edit the same or different file, follow/watch a certain person, as well as have a chat (without requiring copy-paste) about certain piece of code. (both written or via voice)

If something, large enterprises generally don't support smaller and ambiguous licenses. Therefore, if Zed will allow enterprise licensing (ie. via on-prem license server or volume ordering, SSO, whatever) that would increase their adoption quite well...

← PreviousPage 6 of 13Next →