HNHacker News
TopNewBestAskShowJobs

lelandbatey

4,175 karma · joined July 10, 2012

I'm Leland Batey. I make things for the internet. I'm here: http://lelandbatey.com/
submissionscomments
lelandbatey··on Green card seekers must leave U.S. to apply, Trump administration says
You can enter the US on a tourist visa, without any intent to date or meet someone, commiting no fraud, but then encounter someone in the USA, get to know them, and decide to marry that person, and then marry that person. That can happen in 6 months, the length of a tourist visa.

Are you saying that in such cases, the US rules here are and should be that the married couple should live apart for years due to the bylaws of the USCIS?

lelandbatey··on Green card seekers must leave U.S. to apply, Trump administration says
This is only true in the cases for folks on longer visas. If you meet the love of your life and marry them on a tourist visa, you'll be forced to leave your spouse and head back to your country of origin for probably about a year while you wait for USCIS to process I-130.
lelandbatey··on Two EA-18 fighter jets collide at Mountain Home airshow, pilots ejected safely
Sure that's why the bean counters wrote the checks for them, but that's not the reason people attend. People attend because they are a spectacle.
lelandbatey··on Frontier AI has broken the open CTF format
CTF stands for "Capture The Flag" in the parent article. Just the security competition kind, not the FPS game kind.
lelandbatey··on Show HN: Building a web server in assembly to give my life (a lack of) meaning
Specifically how can you use pgx with sqlite while pgx is a postgres-specific library? Sqlc works great with Postgres or Sqlite, Sqlc works with pgx when connecting to Postgres, but pgx can't be used with Sqlite AFAIK
lelandbatey··on I want to live like Costco people
If you're willing to drive far and are in the western US, I highly recommend WinCo foods as a place to buy all your "normal" foods in individual units for very low prices no membership required. Theyre outside the center city usually.
lelandbatey··on ZAYA1-8B matches DeepSeek-R1 on math with less than 1B active parameters
For context, I'm feeling like I have a "free Sonnet" now that I've got Qwen3.6 35B running on my 5070ti at home (I connect to it via Tailscale). I run it _almost exactly_ the same as this Reddit post which found a good way to squeeze the 35B model onto a GPU with 16GB of VRAM: https://www.reddit.com/r/LocalLLaMA/comments/1sor55y/rtx_507... I really like it because it's slightly more operationally complex (I had to write a script to start it) but now that I have it, I literally never have to change it. It's a folder with the llama-server in it and with the model.gguf in it, I run the script which starts serving the model, done.

Like that post, I get 75 tokens/second. The exact model is: Qwen3.6-35B-A3B-UD-Q4_K_M.gguf and I get 128k of context

I run it on my home machine and connect to it from anywhere over tailscale. I connect through the opencode CLI which I configure with this as provider by adding the following to my `~/.config/opencode/opencode.json`:

    {
      "provider": {
        "vllm": {
          "npm": "@ai-sdk/openai-compatible",
          "name": "local-llm-qwen3.6-35B",
          "options": {
            "baseURL": "http://homepc.tail987654.ts.net:8033/v1"
          },
          "models": {
            "Qwen3.6-35B-A3B-UD-Q4_K_M.gguf": {
              "name": "Qwen3.6-35B"
            }
          }
        }
      }
    }
lelandbatey··on Pope Leo called his bank's customer service line. They hung up on him
Sometimes even the pope wants to buy a birthday card for a niece or nephew.
lelandbatey··on Accelerating Gemma 4: faster inference with multi-token prediction drafters
https://chatjimmy.ai being a demo of the "burn the model to an ASIC" approach being sold by Taalas[0], an approach which they use to run Llama 3.1 8B at ~17000 tokens per second.

[0] - https://taalas.com/products/

lelandbatey··on Does Postgres Scale?
Maybe. But as someone who happily self hosting pretty big Temporal workloads for my day job (I inherited it from early adopters circa 2022), it definitely does not feel like a dumpster fire. It chugs along unglamorously and I enjoy working on it.
lelandbatey··on Does Postgres Scale?
The reason that DBOS isn't as popular is because it's younger. DBOS launched in the form we know it in 2024. Temporal is much older; Temporal is technically a fork of Cadence and Cadence released originally in 2017, with Temporal forking and releasing back in 2020. When all three are trying to be "the same sort of thing" and that thing is new, it's hard to show up 7-8 years after the trailblazers and say "oh yeah, we're clearly better" when the existing thing works and is used by tons of folks.
lelandbatey··on How ChatGPT serves ads
Ah yes, the classic "my business plan is your moral problem; you owe me your eyes on my ads because I'm the idiot giving things away for free."

People don't want ads. You imply that "if you accept ads then things will be free" but they will not. Never accept ads. Not for a free service, certainly not in a paid product. Ads exist to enable leaching in both direction in exchange for what ends up being nearly mind control. But it is two-way leaching - companies benefit without the friction of explicit payment, consumers get a service without explicitly paying via money. The downside is neither can stop the bad-incentives motivating bad actions from the other side.

Ads are a deal with the devil, and rejecting them outright is allowed via that deal, just as companies can withdraw their free service. It cuts both ways.

lelandbatey··on Waymo in Portland
The USA is riding on 100 years of the benefits of scale and economic investment. If the USA was investing $150B a year into building passenger rail, we would not be paying $1B/mile for "new" and "trailblazing" rail projects. If we had to build much of the highway infra we built 75 years ago, but do it now, we'd be paying similar $1B/mile prices for the highways.
lelandbatey··on Waymo in Portland
The sad truth is the USA spends ~$150B/year building and maintaining it's road network (to say nothing of the inflation-adjusted costs that went into its initial roll-out). Source: The US Fed tracks it directly - https://fred.stlouisfed.org/series/TLHWYCONS

That's a $41/month subscription every citizen's paying no matter what. When we're pulling cash on that volume from everyone's pockets to build lavish infrastructure literally up to people's doors (vastly more road square footage than housing+school square footage combined), of course folks are going to say "nothing compares" -- because nothing does compare. Which stinks (imagine if we'd focused a century of spending on rail at rates like that; damn), but it is what we have at the moment.

lelandbatey··on A good AGENTS.md is a model upgrade. A bad one is worse than no docs at all
The 100% read rate is very harness/CLI dependent. The "original" idea for AGENTS.md was: the AGENTS.md file will be included as-is in the system prompt by the harness, so the agent doesn't have any choice in whether it'll be read or not. For example, this is a shortened form of what opencode sends as a system prompt for a new session when interacting with a provider (displayed in YAML for formatting, and edited for formatting):

    model: foo-model
    max_tokens: 32000
    top_p: 1
    messages:
      - role: system
        content: |
          You are opencode, an interactive CLI tool that helps users with software engineering tasks.
          Use the instructions below and the tools available to you
          # ... snip ...
          Here is some useful information about the environment you are running in:
          <env>
            Working directory: /home/user/dir
            Workspace root folder: /
            Is directory a git repo: no
            Platform: linux
            Today's date: Tue Apr 28 2026
          </env>
          Skills provide specialized instructions and workflows for specific tasks.
          Use the skill tool to load a skill when a task matches its description.
          No skills are currently available.
          Instructions from: /home/user/dir/AGENTS.md
          # Overview
          This directory holds the entirety of the code for the <dayjob> company. All code lives in Github
          under the `<dayjob>` organization, and beneath that Organization is a wide-and-flat set of all
          the Git repositories of all source code at <dayjob>. That Github repo structure is replicated in
          this directory via `ghorg`.
My AGENTS.md file contents start at the "# Overview" line.

Notice that the harness is just unceremoniously dumping the AGENTS.md file into the exact same text stream as the system prompt, barely contextualizing that hey, starting now, this text is from AGENTS.md and not from the harness.

If you want AGENTS.md to work (likewise, if you want skills or anything else to work) you have to know how the harness is handling/feeding them to the LLM, because no LLM will reliably look on their own.

lelandbatey··on Niri 26.04: Scrollable-tiling Wayland compositor
i3 shows title bars (shrinking or expanding them) kind of like how tabs are displayed in a web browser. If you do nested layouts with i3, though, you wont be able to see all the title bars, but otherwise usually you can see all the title bars (though truncated to fit). That's a pretty common workflow, and it gives you "perfect vision" of all the windows in that workspace. Vs niri which by default scrolls the whole window (title bar included) off the screen, so you can't see all the windows in the workspace at a glance.
lelandbatey··on Netflix Prices Went Up Again – I Bought a DVD Player Instead
Do note that there's a ton of media that is not very available via sailing the seas. For example, reality TV has very low availability, typically only currently airing seasons at best. Thus if you're trying to convince a family to cancel all the big streaming platforms, you may have a hard time when they learn that they can't watch old episodes of "The Real Housewives of the middle of nowhere" from 2012.
lelandbatey··on Understanding the Kalman filter with a simple radar example
Kalman filters are very cool, but when applying them you've got to know that they're not magic. I struggled to apply Kalman Filters for a toy project about ten years ago, because the thing I didn't internalize is that Kalman filters excel at offsetting low-quality data by sampling at a higher rate. You can "retroactively" apply a Kalman filter to a dataset and see some improvement, but you'll only get amazing results if you sample your very-noisy data at a much higher rate than if you were sampling at a "good enough" rate. The higher your sample rate, the better your results will be. In that way, a Kalman filter is something you want to design around, not a "fix all" for data you already have.
lelandbatey··on I quit. The clankers won
> They work

For some definition of work, yes, not every definition. Their product is not without flaw, leaving room at for improvement, and room for improvement by more than only other AI.

> There are no vulnerabilities

That's just not true. There's loads of vulnerabilities, just as there's plenty of vulnerabilities in human written code. Try it, point an AI looking for vulns at the output of an AI that's been through the highest intensity and scrutiny workflow, even code that has already been AI reviewed for vulnerabilities.

lelandbatey··on Netflix raises prices for every subscription tier by up to 12.5 percent
Reading around a bit, yes to Netflix adding anti-piracy measures, maybe to folks recording HDMI/DisplayPort.

Apparently, Netflix is using steganography/content watermarks in their 4k content itself to trace users who are pirating. This is from a totally unsourced Reddit thread[0] but they do reference a real company which claims to do this watermarking[1]. The claim is that in addition to Netflix requiring 4k content to be available only on platforms with Trusted Execution Environments[2], Netflix also encodes each ~10 second "chunk" of the video stream into at least 2 different versions: an Y and a Z version. Then, they serve each customer a unique series of chunks when that customer streams their content, e.g. YYZYZZZYZYYZYZYYZZYZYYZ. Then when content leaks, Netflix can examine each chunk of the leaked content to extract the ID of the user who streamed the content. Apparently, Netflix can encode a lot more than just the userID, they can also encode stuff like the individual device ID, the TEE key ID, etc.

I know you might be thinking "I could do something to defeat that" and you're probably right (e.g. take streams from multiple users and intercut them so that the bits of the watermark through time are being constantly shuffled), but I'll also bet that there's many layers of steganography we don't know about, and unless you get them all, you'll not escape scot-free.

[0] - https://www.reddit.com/r/Piracy/comments/1rqkyjg/with_a_lot_...

[1] - https://irdeto.com/video-entertainment/irdeto-anti-piracy

[2] - https://en.wikipedia.org/wiki/Trusted_execution_environment

lelandbatey··on Flighty Airports
I agree, I find that the "MiseryMap" from flightaware is less "pretty" but much more informationally dense. https://www.flightaware.com/miserymap/
lelandbatey··on Two pilots dead after plane and ground vehicle collide at LaGuardia
Yes, when they work overtime they get paid more for that overtime than regular time.

The money doesn't somehow make it sustainable for the people burning out their lives. Working 7 days a week, including overnight shifts, for 20 years to collect a pension seems like WELL earned compensation.

That's seems unrelated to "we have so few" and "we enmiserate the one's we do have".

lelandbatey··on Chest Fridge (2009)
Mostly yes. Upright fridge and freezer designs trade off efficiency for convenience (rooting around in a chest fridge/freezer can be annoying). https://youtu.be/CGAhWgkKlHI
lelandbatey··on Chest Fridge (2009)
The video down thread shows, the internal food-supports are all wire meshes with big gaps. The cold air is not squirted up and out like a syringe, it's more like the food is kept in a birdcage that's lowered and raised out of a pool of cold air.
lelandbatey··on Delve – Fake Compliance as a Service
Fly is not saying "just ignore SOC2 compliance". Fly is saying "yes, get SOC2, we had to become SOC2 compliant, and also, you can work with your auditor to achieve SOC2 compliance in a more sane way than if you just do whatever is recommended upfront."

Basically, they are saying that you should tailor your SOC2 implementation so that it's actually useful without being a horrible overbearing process, that you have that option and should take it.

lelandbatey··on Warranty Void If Regenerated
Who can know what the world will look like as we "transition"? I sure don't, but I'm thankful the author here has taken a stab at it. I feel like this is one of the first stories I've seen to try to imagine this post-transition world in a way that isn't so gonzo as to be unrelatable. It was so relatable (the human-ness shining all the brighter in a machine-driven world) that I cried as I finished reading. I've felt very anxious about my own future, and to see one possible future painted so vividly, with such human and emotionally focused themes, triggered quite an emotional reaction. I think the feeling was:

> If the world must change, I hope at least we still tell such stories and share how we feel within that change. If so, come what may, that's a future I know I can live in.

lelandbatey··on Where does engineering go? Retreat findings and insights [pdf]
You're right that this isn't some groundbreaking revelation. If you're using AI enough to be feeling it, you're feeling/seeing what they're talking about. The purpose of a paper/retreat like this it get it all together and written down on paper, then to disseminate it to the wider world. I think the paper does a good job of collecting info that isn't wrong, and which has enough info to help guide folks making decisions.
lelandbatey··on “This is not the computer for you”
Don't downvote, it certainly did for me. My first computer was a MBP 13inch from 2009, as I was apple obsessed like the person in the parent article. Time passes and I realized I really didn't like either Windows or Mac, and for the past 10 years Ive been linux only. It really does happen, even if rarely.
lelandbatey··on Whistleblower claims ex-DOGE member says he took Social Security data to new job
Yes, that's the joke
lelandbatey··on Tech jobs are getting demolished in ways not seen since 2008
I think the timelines are too short for trends to be completely apparent yet. You can typically hire people faster than you can scale your income sources, even in the face of tremendous demand. Right this moment there's factors pushing folks to fire, but I also do see some companies delivering more (not a lot, but noticably more) and seeing increasing sales as a result. Those are in conflict, and we'll see which way the trends push through time.
← PreviousPage 3 of 34Next →