Hacker News: bonoboTP

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Fri, 19 Jun 2026 14:11:47 +0000

These are vibes. ViT has been shown to work fine on small data with proper hyperparam and most of what you mention is actually doable just fine with the other architecture as well.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Fri, 19 Jun 2026 14:09:34 +0000

Exactly. Most of the comparison papers are useless. This is hard stuff, only few people have the chops it takes to even attempt this. You can of course train some models and then post the numbers, that's not the hard part.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Fri, 19 Jun 2026 14:07:49 +0000

What's the use case enabled vs running a ConvNeXt or EfficientNetV2 and using the resulting strided features as you would the resulting tokens of a ViT? I'm not saying that ViT is worse. Just saying that the scholarship around comparing them is very bad or nonexistent. You have to properly tune the hyperparam enters on both sides in a fair way, and use all the general modern training tricks also on the CNN side to make it fair.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Fri, 19 Jun 2026 07:40:39 +0000

Yes but not based on rigorous comparison. I'm not saying ViT is bad. But it took over mainly because it's the shiny new thing. It very bandwagon-Y even among PhD students.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Thu, 18 Jun 2026 22:11:31 +0000

There's a dearth of research properly comparing them.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Thu, 18 Jun 2026 22:11:03 +0000

You can run a CNN and use the downsampled feature map the same way as patch tokens.

New comment by bonoboTP in "DeepSeek Introduces Vision"

bonoboTP — Thu, 18 Jun 2026 17:55:43 +0000

CNNs are fine when trained with a good recipe. There are very few good studies comparing them with proper hyperparam search and all the training tricks applied consistently. Transformers are good but ViT vs CNN is not some settled issue. Transformers are more hyped and more popular with the tech enthusiasts who just read forums and news, but if you need stuff done, CNNs are still great.

New comment by bonoboTP in "Microsoft new Outlook takes 10 seconds to do what Outlook Classic does instantly"

bonoboTP — Thu, 18 Jun 2026 14:10:14 +0000

Why are you not on thunderbird yet? Why do you get Windows notifications? Are you using Windows? I don't understand how there are people who can notice such things but still use windows in 2026. Also, please don't write with AI. This post was written with AI.

New comment by bonoboTP in "I found 10k GitHub repositories distributing Trojan malware"

bonoboTP — Thu, 18 Jun 2026 14:08:13 +0000

Exactly. All these ideals work in theory but then in reality banks are also incompetent and will use all kinds of domains.

Same with meta and Google where they often direct you to domains that aren't under their main one and it's actually legit, but there's no way to know. It's impossible to teach family members to pay attention if it's really that domain because it's often legit not that domain.

New comment by bonoboTP in "Automating myself out of development"

bonoboTP — Sun, 14 Jun 2026 17:49:36 +0000

Why can't it do that? How do you do it then? An LLM-based agent can execute any commands in a terminal, including running whatever commands that will send commands to an external device through whatever wires you have set up. Is the claim that you need to have hands-on access to fiddle with the embedded device? That there is no way to just send wire signals out from the workstation towards the device that could load the program, see the results etc?

New comment by bonoboTP in "Automating myself out of development"

bonoboTP — Sun, 14 Jun 2026 12:35:03 +0000

Ok but it takes weeks of trial and error for expert humans. If you allow the llm to run the code and see the results and do trial and error, it will also likely figure it out, no?

New comment by bonoboTP in "Statement on US government directive to suspend access to Fable 5 and Mythos 5"

bonoboTP — Sat, 13 Jun 2026 23:33:32 +0000

Yes. We will likely have a heavily policed narrow line to some kind of AI during the phase when we as employees will still have value in this centaur configuration, so we'll be allowed to steer the largely autonomous models for our employers, direct them a bit when they do something wrong, if we can convince the guard-AI. All this will be logged and constantly audited by your employer and by the government and the AI companies.

For hobby use, only simple models will be allowed. The models of the future will be so large that the huge data centers being built right now will be necessary for training them. Open sourcing will be a strategic nation state level decision. Regular people won't have it. We will have toys, we will have infinite AI generated entertainment and the rest isn't exactly planned out I guess.

New comment by bonoboTP in "Automating myself out of development"

bonoboTP — Sat, 13 Jun 2026 21:20:28 +0000

Did you try this by giving it access to the materials? Human programmers also don't memorize all this stuff. If this is the reason for your calmness it's quite shortsighted.

There are problems when you rely too much on AI generated code, but these shallow dismissals are quite annoying.

New comment by bonoboTP in "Claude Fable 5"

bonoboTP — Wed, 10 Jun 2026 07:15:02 +0000

Meta made Pytorch and a lot of vision models back in the day, like Faster RCNN, Mask RCNN, the Detectron framework, and more recently the SAM and DINO series. AI not just LLMs.

New comment by bonoboTP in "Home alone: Remote work, isolation, and mental health"

bonoboTP — Tue, 09 Jun 2026 16:48:18 +0000

Yes, you need a happy medium. I know complainers who do zero socialization, never stay for social events even once per month, never join a Friday beer (sure, you can drink alcohol free) etc then complain they are overlooked. Humans are humans. Work and life is an artificial separation from the last few decades. It was never a concept before.

I'm not talking about doing all your hobbies in the office and staying 4 hours more than your official hours just shmoozing every day.

Working effectively with people requires some level of personal connection, whether that's shared lunches, chatting over coffee etc. Some just want to plug their earbud in and clock out at 5, and insist to only spend time by typing code in the IDE. I'm saying that this won't work, no matter how they complain online. Humans are social, you have to deal with it.

New comment by bonoboTP in "OpenCV 5 Is Here: The Biggest Leap in Years for Computer Vision"

bonoboTP — Tue, 09 Jun 2026 16:39:44 +0000

By doing it. Decide on a small project, like tracking your cat, detecting food items in your fridge, then take it step by step.

Then do a slightly more ambitious project. Start with something very simple.

It also heavily depends on what you already know regarding programming, image processing etc.

New comment by bonoboTP in "How LLMs work"

bonoboTP — Sun, 07 Jun 2026 19:54:17 +0000

Yes, but it wasn't invented from nothing in 2017. Soft attention existed in other applications like information retrieval, Nonlocal networks had similar ideas as well. But it wasn't seen or used as a fundamental building block. But it wasn't something out of the blue either.

New comment by bonoboTP in "Home alone: Remote work, isolation, and mental health"

bonoboTP — Sun, 07 Jun 2026 17:13:49 +0000

I didn't realize that was what you're thinking about. Well yeah, those guys invest in social connections at work, while you are not. Alliances, trust and human relationships will always matter, no matter how much people complain about this online. It will always be a benefit to be liked and known personally.

New comment by bonoboTP in "LLMs are eroding my software engineering career and I don't know what to do"

bonoboTP — Sun, 07 Jun 2026 17:09:58 +0000

That's the right question. I don't like this dichotomy between domain and engineering. It seems to come from people who just build different CRUD apps for mobile and websites for businesses in different industries and that's what they call domain.

Not like a webdev entering game engine design or a database engineer entering computer vision research, or someone working in embedded hard-realtime systems switching to making video editing GUIs.

New comment by bonoboTP in "LLMs are eroding my software engineering career and I don't know what to do"

bonoboTP — Sun, 07 Jun 2026 16:34:05 +0000

True of human-made things as well. Most video essays don't get more than a dozen views, most gameplay streams similarly. People playing their guitar and uploading, same. SoundCloud, YouTube, twitch. Human-made app store apps is the same story. Most are not downloaded by even 100 people. Most Github repos don't even get a handful of stars.