HNHacker News
TopNewBestAskShowJobs

bensyverson

1,491 karma · joined February 21, 2026

Developer for 35 years, 10 years at IDEO as a Software Designer, project lead and Design Director. Currently working on agentic native apps, and on the hunt for a profitable SaaS company to acquire via infinitecake.net
submissionscomments
bensyverson··on Automating my 35mm film scanning pipeline
Lots of reasons, but ones that matter to me: the batteries last for months, there’s no boot up time, and there’s no screen, which encourages you to keep your own eyes open.
bensyverson··on We want you to build the next Git platform on Cloudflare
Centralization is antithetical to the whole idea of the internet
bensyverson··on Sonnet 5.5
Not sure why I'm being downvoted for stating the obvious. To the parallel agent skeptics: I was also a skeptic until a month or two ago. I would run one agent, watch it carefully, and check its work. However the models got good enough that it was more efficient to do more work in parallel, then have a single agent integrate the changes with a critical eye, then run a code quality pass, and then I would take a look a it and kick the tires behaviorally.

It does involve letting go and not micromanaging every code convention and implementation detail, but that is the same skill you need when leading engineering teams.

bensyverson··on Sonnet 5.5
I have conversations with a smart model, and then the model writes the spec. I review and approve the spec, and it dispatches.

For a concrete example, check out this random plan [0]. A detailed spec followed by the exact implementation tasks that will be executed by the subagents.

[0]: https://github.com/bensyverson/woodcase/blob/main/project/20...

bensyverson··on Sonnet 5.5
I create large hierarchical plans, and have a coordinator agent divvy up the work. It's extremely effective.
bensyverson··on Sonnet 5.5
If you have well-specified tasks, you can easily reach 10 simultaneous agents working on disjoint parts of the code in worktrees. That consumes tokens pretty quickly!
bensyverson··on Floci: Locally emulating any cloud service
I've looked into developing "digital twins" of APIs that I rely on, similar to the StrongDM pattern, but the models talk me out of it every time. The rationale is that there could be subtle differences or drift between the behavior of the real service and the mocked API. So you may think you're on solid ground, but once you switch to prod, you could see show-stopping differences.

I think if I was doing evals on the way an agent interacts with a service (e.g. Slack) the digital twin could make sense, but for rapid iteration or testing, it might not be what you want.

bensyverson··on Plan mode is dead
I use Jobs [0] to manage this—it's an agent-first CLI to track issues and tasks. A single `job orient` command gives the agent the current task in the context of the larger plan. It's a replacement for Plan Mode and issue trackers, and it has allowed me to execute massive plans in parallel with minimal oversight. There's a web UI, but it's a work in progress.

[0]: https://github.com/bensyverson/jobs

bensyverson··on A font that reads what you wrote
I assume it works perfectly in Chrome, the only browser in the world
bensyverson··on Flock is rolling out a voluntary severance program
It's really interesting to me that Flock was the straw that broke the surveillance camel's back, and not, I don't know, the home assistant devices from trillion-dollar advertising and e-commerce companies that listen to everything in your home.
bensyverson··on I built non-autoregressive decision models with RL a year ago
It's either a small language model or a large language model (LLM). It's not a generative model, but neither is BERT, which is also a language model.
bensyverson··on I built non-autoregressive decision models with RL a year ago
Breaking news: small language models struggle with math
bensyverson··on I built non-autoregressive decision models with RL a year ago
I guess we’ve just reached the point where everyone has to state the obvious, and common sense is extremely uncommon.

So here goes: you should not use an AI model to validate a claim which is trivial to calculate deterministically. That is (obviously?) not what a model like Jev is for, thus it is not a good test of Jev.

bensyverson··on I built non-autoregressive decision models with RL a year ago
This is not a good faith test of the system.
bensyverson··on I built non-autoregressive decision models with RL a year ago
And yet no one cared about this research until it was productized and communicated well. Multitouch existed before the iPhone.
bensyverson··on We made Playwright 2x faster and 80% more token efficient
WebKit is available on Linux, just not Apple’s wrapper. It would certainly be possible to give Sleepy a Linux backend if you’re motivated!
bensyverson··on We made Playwright 2x faster and 80% more token efficient
If your needs are simpler, I created a tiny headless WebKit browser specifically for agents called Sleepy Hollow [0]

[0]: https://github.com/bensyverson/sleepyhollow

bensyverson··on CC is an AI agent for families and groups
Interesting choice with the name. Claude Code is sometimes abbreviated to CC.
bensyverson··on Pion, an agent designed to run any company autonomously
I’d love to see your pass at a “unique and interesting” skill. I’ve been working on tools and techniques in this area, and it’s harder than you might think.
bensyverson··on Base84 deserves a place in file names
The non-alphanumeric characters are just asking for trouble. Agents will fail to escape them correctly, and the $ could expand a variable.

Base62 (or 64, with -_) is totally sufficient.

bensyverson··on HuggingFace: Security.txt
Looks about as effective as Robots.txt
bensyverson··on Ask HN: Can we please limit the AI news flood?
Yeah, that's like someone in 1998 saying we're at peak internet hype
bensyverson··on Apple A20 Pro Geekbench Result
And 20% YOY compounds quickly… If the speed doubles in 4 years and the RAM situation gets better, we could be looking at some very capable on-device models in 2030.
bensyverson··on Creativity is the new moat
The text is clearly not written by AI… people need to chill on this accusation. Yes, the blog design itself looks Claudish, but if you actually read the post, no AI writes like that.
bensyverson··on Thanks to Siri Recaps, your Apple Watch is always listening
Our devices are always listening, if you have them set up to listen for hail words like "Siri" or "Hey Google," so I think the Overton window on listening already shifted about 10 years ago. The device actually recording/transcribing what it hears is newer, but you're right that it's becoming more and more normalized. With Granola, I think most people assume that any video call is being transcribed, even if there's no indication in the meeting.

Whatever side you come down on, privacy doesn't stand a chance against a small amount of convenience.

bensyverson··on Show HN: The same nine streaming subscriptions cost $702/year more than in 2021
The stock /frontend-design skill really loves an all-caps mono caption
bensyverson··on DaVinci Resolve 21.1
Not anymore! [0]

[0]: https://www.mturk.com

bensyverson··on DaVinci Resolve 21.1
Completely agree. How many people over the past 20 years have been musically gifted, but found ProTools or Logic impenetrable? The AI unlock is about enabling that creativity, not replacing it.
bensyverson··on DaVinci Resolve 21.1
It's already possible, but in my opinion it will never be popular. In some ways this has already been proven through other media; ChatGPT can easily generate a novel, but how many people are reading pure-AI novels?
bensyverson··on I tested 10 model/harness combinations on the same Three.js task
Yes, I'd be really interested to see Fable / Opus / Sonnet.

GLM 5.3 Flash Max had an interesting showing. Its Codex version was bad [0], and it completed in 9 minutes. The OpenCode version was much richer [1] and more detailed, completed in 20 minutes. And the OMP version was arguably the most complete [2], completing in 30 minutes.

This is probably the strongest argument for the effect of a harness, and I'd be interested to learn the differences in the prompts and tools between these three.

[0]: https://alvins82.github.io/hangar-harness-model-tests/hangar...

[1]: https://alvins82.github.io/hangar-harness-model-tests/hangar...

[2]: https://alvins82.github.io/hangar-harness-model-tests/hangar...

← PreviousPage 3 of 14Next →