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 Google Engineers Launch "Sashiko" for Agentic AI Code Review of the Linux Kernel
To be honest, during my ~10 years of experience, I haven't stumbled on any _project_ that has _sufficient_ amount of either testing, static analysis, and/or, where everyone who contributes gives a damn about linter/formatter warnings or errors.

Unless there is a CI that fails immediately when these checks fail, and it is not possible to bypass/override, then of course quality in the codebase improves significantly.

But that is very hard to give business results, in terms of ROI, 99.9% of the PMs will sweep these under the rug, because of unclear ROI.

Another remark I have to make is that, these things should be in a cloud/container (isolated environment) somewhere. Because I know several people who has their git commit aliased with no-precommit-hooks, because well, otherwise they complain that it is too slow. (Slowness is another issue that can be improved, but again, less visible ROI than just adding the command-line switch)

pvtmert··on macOS 26 breaks custom DNS settings including .internal
I am pretty sure this will get fixed as Amazon had depended on this feature to resolve internal domains through the ACME managed openvpn-service. Rather than overriding global DNS settings which may break on certain hotspot configurations.
pvtmert··on MacBook Pro with M5 Pro and M5 Max
Nope, they are required to have an option to opt-out from adapter. They choose to charge for one!

https://9to5mac.com/2025/10/16/no-the-eu-didnt-ban-apple-fro...

pvtmert··on MacBook Pro with M5 Pro and M5 Max
I know EU didn't ban chargers, but the common American sentiment somehow molded into that.

It is interesting to see how mass-propaganda is playing out right before our eyes...

pvtmert··on New accounts on HN more likely to use em-dashes
Having mixed feelings on word "actually" as it is/was one of my favorites. Other stuff like "for instance" and "interestingly" are seem to be getting there too...
pvtmert··on New accounts on HN more likely to use em-dashes
Honestly, first paragraph sounds more human and sincere for sure.

Also adding better "context" into the discussion, than the usual claims/punchlines of marketing-speak.

Maybe it's not exactly the grammar itself but also overall structuring of the idea/thought into the process. The regular output sounds much more like marketing-piece or news-coverage than an individual anyway. I think, people wanna discuss things with people, not with a news-editor.

pvtmert··on New accounts on HN more likely to use em-dashes
I started making deliberate grammar and spelling mistakes in professional context. Not like I have a perfect writing anyway, but at least I could prove that it was self-written, not an auto-generated slop. (Could be self-written slop though :)

This applies not only work-stuff itself also to the job-applications/cv/resume and cover-letters.

pvtmert··on Bus stop balancing is fast, cheap, and effective
Imagine the delays are so prominent, someone decides to make a website for CTA (call-to-action) and semi-regularly shares updates on it...

I've been to Seattle once, (ex-Amazon here) where the DevCon was held in the town while my team was located in Bellevue. I took initiative to rent a bike for a day (60$ for drop-bar gravel bike) I must say although I did not beat the time between Day-1 (Office across spheres) and Bingo (Bellevue office), it was not far off. Even comparing the "Shuttles" Amazon operated, shuttle took about 1h while ride takes around 1h15m. (Plus sweat)

> P.S: I would say I am in a "fair" shape as I ride quite a lot throughout the year.

pvtmert··on SSH Virtual Hosting: No Host Header, but Public Key
I was checking some of the bookmarks/reading-lists I had earlier, stumbled on this piece on the exe.dev's blog.
pvtmert··on A simple web we own
Apache doesn't have it on by default but easy to turn it on. It's called usermod or mod_user. By default it's the ~/www directory. So, anyone with /home/<name>/www ends up being site.url/~<name>/

It is also possible to add .htaccess and other things there, like username/password challenge (WWW-Authenticate) into that on per-user basis.

Mostly universities had hosting setup the same way. ISPs would also offer a similar thing with an additional fee to your internet-subscription. They mostly provided FTP to upload files. Nowadays if anyone tries to, it will be a SFTP rather than FTP.

pvtmert··on Attention Media ≠ Social Networks
Unrelated to the topic described in the blog itself, I overall like the theme of `susam.net`. The name itself reminded me of a sesame seed in Turkish for a while. (I think author had recently mentioned one of the recent posts that they wanted to get susam.com but that was already taken by a Turkish company selling some spices...)

The content (that shows up in HN) is also good. Since I am on mobile device, I cannot tell the exact font used, but seems like Georgia to me. While https://github.com/susam/susam.net hosts the actual source code of the website.

Another remark: Would be really nice to have a same theme adaptation for BearBlog and similar places.

pvtmert··on Carelessness versus craftsmanship in cryptography
From the post itself, I am not sure if the author had sent a patch or some sort of a pull-request to the affected entities. Namely pyaes and aes-js.

The response might've been different if the author had already given a patch, in somewhat backward-compatible way. This doesn't even have to be a functional patch, could be a simple `@warning: usage of default IV will cause insecure storage` similar annotations on the affected functions.

Another thing to remark (and which might've been off-putting for the authors of these libraries) that the author had used term mistakes in various places. Of course in an ideal world, ego should not or would not matter, but these libraries both seem to be quite stale and possibly the authors are having other $DAYJOB responsibilities. Making it difficult to fix things that they just receive complaints about. (I am also guessing these are quite many...)

Again in relation to the points above, it might've been better to say: Cryptography evolves over time, last years' best-practices get outdated, vulnerabilities being found, replaced with newer best-practices of this year. Same will happen next year too. It's not a deliberate mistake or any type of incompetency issue, this is a matter of ever-evolving field that we know and understand better...

pvtmert··on Blue light filters don't work – controlling total luminance is a better bet
My overall take (elephant in the room): Blue light filters don't work, it depends on what you do & how you do it.

For example, most people keep watching/scrolling Instagram Reels and TikTok videos. They keep stimulating the brain constantly, not just at electrical level but also in emotional/chemical level too.

I have seen people who are addicted and cannot get rid of the addiction. This is not only the dopamine-boost, it has deeper connections of neuro-chemical stimuli. Just observe around you; people pick up their phone to directly open Insta/TikTok, start scrolling right away every 5-10 seconds. (watching stories included too)

This is to some extent that when you mention even the possibility of such addiction and abnormal behavior, one gets outright resistance and denial of addiction itself. Much like substance abuse...

My point is, majority of the population watches/scrolls these, needing 10g of melatonin to fall asleep.

Obviously if I get engaged in an interesting stuff continuously, the existence of blue light does not matter that much. It matters if/when I am reading a novel which is in a mediocre chapter where nothing that interesting going on. The existence of blue-light or lack thereof may tip the scale at that point.

pvtmert··on Blue light filters don't work – controlling total luminance is a better bet
Interesting take for me is that melatonin (over-)usage can be severely harmful for the individuals.

  ... over-the-counter melatonin supplements can contain anywhere between 10 to 30 times as much melatonin as is optimal to maintain circadian hygiene. If you have ever taken melatonin and got immediately knocked out cold, had weird dreams and woke up in the middle of the night sweaty or shivering, you likely took too much—which, to be clear, is not your fault, it’s the default in the US and Canada. The mega-doses in stores serve as hypnotics (punches you to sleep), but wreck sleep architecture. The right dose is ~0.3 mg, which is hard to find in pharmacies but can be found online.
pvtmert··on OpenScan
Wouldn't it be much move viable (and accurate) using an iPhone with LiDAR? Recent models like iPhone 16 Pro are also coming with a USB Type-C port, can deliver some watts which is enough to drive a simple motor on a rotational axis. Significantly making overall hardware dumb enough that it will be much cheaper. (ie. Something around 10-20$ range for the rotary-station)

  Could've been possible and even easier with headphone jack too.
pvtmert··on Claws are now a new layer on top of LLM agents
Great idea, happy to ~steal~ be inspired by.

I propose a few other common elements:

1. Another AI agent (actually bunch of folks in a 3rd-world country) to gatekeep/check select input/outputs for data leaks.

2. Using advanced network isolation techniques (read: bunch of iptables rules and security groups) to limit possible data exfiltration.

  This would actually be nice, as the agent for whatsapp would run in a separate entity with limited network access to only whatsapp's IP ranges...
3. Advanced orchestration engine (read: crontab & bunch of shell scripts) that are provided as 1st-party components to automate day-to-day stuff.

  Possibly like IFTTT/Zapier/etc. like integration, where you drag/drop objectives/tasks in a *declarative* format and the agent(s) figure out the rest...
pvtmert··on Claws are now a new layer on top of LLM agents
From my experience, it depends on how you frame your "service" to the reviewers. Obviously 2023 was the very early stage of LLMs, where the security aspects were quite murky at best. They (reviewers) probably did not had any runbook or review criteria at that time.

If you had advertised this as a "regular service which happens to use LLM for some specific functions" and the "output is rigorously validated and logged", I am pretty sure you would get a green-light.

This is because their concern is data-privacy and security. Not because they care or the company actually cares, but because fines of non-compliance are quite high and have greater visibility if things go wrong.

pvtmert··on Claws are now a new layer on top of LLM agents
I am also ex-FAANG (recently departed), while I partially agree the "policy-people" pop-up fairly often, my experience is more on the inadequate checks side.

Though with the recent layoffs and stuff, the security in Amazon was getting better. Even the best-practices for IAM policies that was the norm in 2018, is just getting enforced by 2025.

Since I had a background of infosec, it always confused me how normal it was to give/grant overly permissive policies to basically anything. Even opening ports to worldwide (0.0.0.0/0) had just been a significant issue in 2024, still, you can easily get away with by the time the scanner finds your host/policy/configuration...

Although nearly all AWS accounts managed by Conduit (internal AWS Account Creation and Management Service), the "magic-team" had many "account-containers" to make all these child/service accounts joining into a parent "organization-account". By the time I left, the "organization-account" had no restrictive policies set, it is up to the developers to secure their resources. (like S3 buckets & their policies)

So, I don't think the policy folks are overall wrong. In the best case scenario, they do not need to exist in the first place! As the enforcement should be done to ensure security. But that always has an exception somewhere in someone's workflow.

pvtmert··on Claws are now a new layer on top of LLM agents
Does one really need to _buy_ a completely new desktop hardware (ie. mac mini) to _run_ a simple request/response program?

Excluding the fact that you can run LLMs via ollama or similar directly on the device, but that will not have a very good token/s speed as far as I can guess...

pvtmert··on Single vaccine could protect against all coughs, colds and flus
I would be interested to see such thing on my Twitter/X feed, is it @_in_mice? It seems to be a bit stale. Would you like to share? Thanks!
pvtmert··on Show HN: Git Diff Viewer and Editor
I have couple of questions and suggestions;

Q1: Can this be a replacement/drop-in for $GIT_EDITOR?

Q2: On the issue ID #1, it is already mentioned but, can it use the existing $EDITOR / $VISUAL ?

Q3: Feature request / differentiating factor: Many similar tools exist, but if you can/could create a "snapshot" feature, that may be the killer factor here. As I see this is already a plain-C code, _tar_'ing `.git` directory and restoring after; essentially doing-bunch-of-things-but-cancelling-them-after use-case.

S1: Bit nit-picky but all strings are hard-coded within various places. Although it's easier to find where is something, will be difficult to translate for different languages.

S2: I think the title may need CLI/TUI in it. For some reason I expected yet-another Electron-based app. Maybe it's just me though...

S3: Adding to Homebrew (at least as a CASK even) should be relatively straightforward. As I (usually) don't want to manually maintain some binaries I downloaded from random-github-releases-link. Not to mention auto-updates are handled by the Brew.sh itself...

pvtmert··on Co-Routines in 1-page of C (2013)
I wonder if/when ClaudeCode or others are able to implement such a simple/useful thing in a concise way. Or would they bring hundreds of lines of C...
pvtmert··on Recoverable and Irrecoverable Decisions
The main difference being the time it takes to recover/reverse the decision.

Second point is: You don't need to reverse the decision you took, instead you may find a way to fix the impact but not the root-cause.

It's like when one fucks up the MySQL replication and the data consistency is corrupted. One can manually (and slowly) fix the inconsistency with downtime. Or, spin up a whole new cluster from an existing well-known node/state. Some entities may be missing, but you could gradually add them back later.

Not a reversible, but recoverable decision.

Amazon goes by with one-way vs two-way door decisions internally. Sometimes adding much bureaucracy to the equation. Just-do-it/Bias-for-action aspect usually don't go as far as the recovery period prolongs.

pvtmert··on Recoverable and Irrecoverable Decisions
(delayed)

Only difference is time. Much like an eventually consistent transactions, recoverable decisions have propagation latency.

The breaking part here is that will you able to survive until the recovery is complete?

pvtmert··on Improving 15 LLMs at Coding in One Afternoon. Only the Harness Changed
I am in the same boat. I have built bunch of bash/shell scripts in a folder back in 2022/2023. When models first came out, I would prompt them to use subshell syntax to call commands (ie: '$(...)' format)

I would run it via calling AWS Bedrock API through AWS-CLI. Self iterating and simple. All execution history directly embedded within.

Soon after, I wrote a help switch/command to each script. Such that they act as like MCP. To this day, they outperform any prompts one can make.

pvtmert··on A "QuitGPT" campaign is urging people to cancel their ChatGPT subscriptions
The most ironic bit here is that MIT Tech review shows annoying pop-ups that say "Get 25% off for AI digest!"

Well played!

pvtmert··on Stories from 25 Years of Software Development
Probably they mentioned something like: "Not possible with current hardware speed" just to be translated as "impossible" since the recollection is second-hand.
pvtmert··on LLMs as the new high level language
Except that the output depends on stars' alignment.

Imagine a machine that does the job sometimes but fails on some other times. Wonderful isn't it?

pvtmert··on You Are Here
Honestly, "vibe-code-cleaners" are already out there and in demand!

I even expect "vibe-code-scalers" will come soon, to be able to fix and scale up the spaghetti AI agents plopped in the first place.

The author seems to be an Amazonian, it also seems that they are good at "Invent", but not at "Simplify" bit.

Big-Tech has invented LLMs, that is great. Big-Tech hasn't been great when it comes to "Simplify"-ing things. Actually, notoriously bad at it.

That is the opportunity here; "Simplifying" these workflows, making AaaS (Agent as a Service or AI as a Service)

pvtmert··on You Are Here
> The cost of turning written business logic into code has dropped to zero

It hasn't. Large enterprises currently footing the bill, essentially subsidizing AI for now.

I constantly see comparisons between the 200$ Claude-Code Max subscription vs 6-figure (100k$) salary of an engineer.

The comparison here is, first of all, not apples-to-apples. Let's correct CC subscription to the yearly amount first; 12x200=2400$. Still more than 10x difference compared to the human engineer.

Although when you have the human engineer, you also pay for the experience, responsibility, and you somewhat transfer liability (especially when regulations come into play)

Moreover, creation by a human engineer, unless stolen IP or was plagiarized, owned by you as the employer/company. Meanwhile, whatever AI generated is guaranteed to be somewhat plagiarized in the first place. The ownership of the IP is questionable.

This is like when a layman complains when the electrician comes to their house, identifies the breaker problem, replaces the breaker which costs 5$ and charges 100$ for 10-minute job. Which is complete underestimation of skill, experience, and safety. A wrong one may cause constant circuit-breaks, causing malfunction in multitude of electronic devices in the household. Or worse, may cause a fire. When you think you paid 100$ for 10-minutes, in fact it was years of education, exams, certification, and experience you had paid for your future safety.

The same principle applies to the AI. It seems like it had accumulated more and more experience, but failing at the first prompt-injection. It seems like getting better at benchmarks because they are now part of their dataset. All these are hidden-costs 99% does not talk about. All these hidden costs are liabilities.

You may save an engineer's yearly salary today, at the cost of losing ten times more to the civil-lawsuits tomorrow. (Of course, depending on a field/business)

If your business was not that critical to get a civil-lawsuit in the first place, then you probably didn't needed to hire an engineer yourself. You could hire an agency/contractor to do that in much cheaper way, while still sharing liability...

← PreviousPage 3 of 13Next →