I've tried the latest and greatest agentic CLI and toolings with the public SOTA models.
I think this is a productivity jump equivalent to maybe punch cards -> compiled languages, and that's it. Something like a 40% increase, but nowhere close to exponential.
They're simply bluffing, and you called them on it. Thanks for your service. Too many people think they can just bullshit and bluff their way along and need to be taken down a peg, or for repeat offenders, shunned and ostracized.
That's jump if you are a junior. It falls down hard for the seniors doing more complex stuff.
I'm also reminding that we tried whole "make it look like human language" with COBOL and it turned out that language wasn't a bottleneck, the ability of people to specify exactly what they want was the bottleneck. Once you have exact spec, even writing code on your own isn't all that hard but extracting that from stakeolders have always been the harder part of the programming.
Except punch cards are a data storage format, not a language. Some of the the earliest computers were programmed by plugboard ( https://en.wikipedia.org/wiki/Plugboard#Early_computers ) so that might be thought of as a precursor to machine language / assembly language.
And compiled and interpreted languages evolved alongside each other in the 1950s-1970s.
I used the early web. I miss forums, I miss the small webmaster, I miss making fun, small websites to share with friends.
And while you could make the argument that these forms of media were superior to TikTok, I’d also argue that this is mostly just taste.
While we have closed ecosystems now, they’re much easier to make and share content to than the web of the past. It’s much easier to get distribution and go viral. There’s also a well trodden path to monetization so that if you craft great content people love, you can make a living from it.
Yeah quirky designs, guestbooks, affiliate badges, page counters, all that stuff. I miss it. But only ever a very small fraction of society was going to be able to make and consume that stuff.
This new internet is much more accessible and it occasionally produces diamonds of culture, you just have to know where to look.
So no, I don’t think any amount of decentralized protocols or tooling or any technology really can change this. I think this trend is set and will continue, and I’ve had to learn to be more open minded to how I perceive internet content.
No one is going to make personal websites or change their behavior in a major way.
Look, you can still sign up for free web hosting and make an HTML page and tell your friends. There are still people that do this. But it’s naturally eclipsed by these other methods of much easier content sharing.
The point is the content itself, not the packaging. Just get over the shape of the packaging and enjoy.
Precisely. I have made my own e-cards to send to friends to commemorate holidays and outings. All HTML + CSS, responsive and looks fine on all devices.
> I used the early web. I miss forums, I miss the small webmaster, I miss making fun, small websites to share with friends.
None of these things are gone. They're just not new anymore for a lot more people, and they probably have significantly less social impact and cachet. But that's all.
Yeah. It's actually the opposite: the original web considering of home pages and niche forums participated in exclusively by actual humans is very much still there, even if sometimes in different places.
It's "web 2.0" consisting of centralised networks full of your friends and friends of friends posting photos, updates and invitations that's being killed by those networks promoting "engagement bait" and generated content and bot accounts
You spent 3 months on this hacked together garbage when you probably could’ve just configured a pre-existing solution off the shelf with like 10 minutes of reading and understanding documentation.
This blog post reeks of “you can just do things” type of engineering. This is the quality of engineering I would expect from “TPOT” (that part of Twitter) where people talk about working 12 hour days. It’s cause they’re working 12 hours on bullshit like this.
Building some sweet custom codec or binary transportation algorithm was barely cute in like 1989. It definitely ain’t cute now.
How many of these AI and “agentic” companies are just misled engineers thinking they are cracked and writing needlessly complex solutions to problems that dont even exist?
The year is 2034. Countless attempts at re-producing the sophisticated wetware of the brain have failed. Modeling research has proved unfruitful, with the curse of dimensionality afflicting every attempt at breaking the walls of general intelligence. With only a few million of capital left, and facing bankruptcy, they knew that only one option remained.
It's becoming challenging to really evaluate models.
The amount of intelligence that you can display within a single prompt, the riddles, the puzzles, they've all been solved or are mostly trivial to reasoners.
Now you have to drive a model for a few days to really get a decent understanding of how good it really is. In my experience, while Sonnet/Opus may not have always been leading on benchmarks, they have always *felt* the best to me, but it's hard to put into words why exactly I feel that way, but I can just feel it.
The way you can just feel when someone you're having a conversation with is deeply understanding you, somewhat understanding you, or maybe not understanding at all. But you don't have a quantifiable metric for this.
This is a strange, weird territory, and I don't know the path forward. We know we're definitely not at AGI.
And we know if you use these models for long-horizon tasks they fail at some point and just go off the rails.
I've tried using Codex with max reasoning for doing PRs and gotten laughable results too many times, but Codex with Max reasoning is apparently near-SOTA on code. And to be fair, Claude Code/Opus is also sometimes equally as bad at doing these types of "implement idea in big codebase, make changes too many files, still pass tests" type of tasks.
Is the solution that we start to evaluate LLMs on more long-horizon tasks? I think to some degree this was the spirit of SWE Verified right? But even that is being saturated now.
Totally agree. I just got a free trial month I guess to try to bring me back to chatGPT but I don't really know what to ask it to display if it is on par with Gemini.
I really have a sinking feel right now actually of what an absolute giant waste of capital all this is.
I am glad for all the venture capital behind all this to subsidize my intellectual noodlings on a super computer but my god what have we done?
This is so much fun but this doesn't feel like we are getting closer to "AGI" after using Gemini for about 100 hours or so now. The first day maybe but not now when you see how off it can still be all the time.
The good old "benchmarks just keep saturating" problem.
Anthropic is genuinely one of the top companies in the field, and for a reason. Opus consistently punches above its weight, and this is only in part due to the lack of OpenAI's atrocious personality tuning.
Yes, the next stop for AI is: increasing task length horizon, improving agentic behavior. The "raw general intelligence" component in bleeding edge LLMs is far outpacing the "executive function", clearly.
Shouldn't the next stop be to improve general accuracy, which is what these tools have struggled with since their inception? Until when are "AI" companies going to offload the responsibility on the user to verify the output of their tools?
Optimizing for benchmark scores, which are highly gamed to begin with, by throwing more resources at this problem is exceedingly tiring. Surely they must've noticed the performance plateau and diminishing returns of this approach by now, yet every new announcement is the same.
What "performance plateau"? The "plateau" disappears the moment you get harder unsaturated benchmarks.
It's getting more and more challenging to do that - just not because the models don't improve. Quite the opposite.
Framing "improve general accuracy" as "something no one is doing" is really weird too.
You need "general accuracy" for agentic behavior to work at all. If you have a simple ten step plan, and each step has a 50% chance of an unrecoverable failure, then your plan is fucked, full stop. To advance on those benchmarks, the LLM has to fail less and recover better.
Hallucinations is a "solvable but very hard to solve" problem. Considerable progress is being made on it, but if there's "this one weird trick" that deletes hallucinations, then we sure didn't find it yet. Humans get a body of meta-knowledge for free, which lets them dodge hallucinations decently well (not perfectly) if they want to. LLMs get pathetic crumbs of meta-knowledge and little skill in using it. Room for improvement, but, not trivial to improve.
I stopped listening to Lex Fridman after he tried to arbiter a "peace agreement" between Russia and Ukraine and claimed he just wanted to make the world "love" each other more.
Then I found out he was a fraud that had no academic connection to MIT other than working there as an IC.
> I stopped listening to Lex Fridman after he tried to arbiter a "peace agreement" between Russia and Ukraine...
Same here. I lost all respect for Lex after seeing him interview Zelensky of Ukraine. Lex grew up in Moscow. He sometimes shows a soft spot for Russia perhaps because of it.
USB is fine for PS1/retro games; should be more than enough for AthenaEnv. The difference only matters for PS2 backups. And there're more options than those two. HDD/SSD, Ethernet, MX4SIO/SIO2SD, MMCE (SD2PSX et al).
That's true; network too. Can play digital backups (off USB/Eth) only by using POPStarter (for those unaware, POPS being Sony's PS1 emulator ripped off the single game that was officially used on). Although POPS isn't really that good (was used, experimentally, only once afterall), USB throughput isn't an issue.
FreeMcBoot on a memory card + USB flash drive (or internal disk) is a popular option to play on real hardware. Saves wear and tear on the optical drive too.
It's a good time! I feel the PS2 is an iconic member of any living room TV setup, especially with an SSD, a couple controllers, and component out. Nice to have for hangouts!
The original Phat PS2 models support internal IDE disks, though it's easy to convert them to SATA with an inexpensive adapter. Then prepare the SSD and install (I use WinHIIP, "obsolete" but works fine). It noticably improves startup and load times in games!
Some also swap the fan for a Noctua to give it a complete "quiet upgrade", but my stock fan isn't very loud.
I’ve felt the same. Also the AGI outcome for software engineers is:
A) In 5 years no real improvement, AI bubble pops, most of us are laid off.
B) In 5 years near—AGI replaces most software engineers, most of us are laid off.
Woohoo. Lose-lose scenario! No matter how you imagine this AI bubble playing out, the musics going to stop eventually.
Kids music toys are often just purely toys tap a button, make a sound... But the skill ceiling could be so much higher, offering the ability to learn and express themselves more. Awesome work.
For anyone who likes the idea of better music "toys" for their kids (i.e. not toys at all, but still easy to use) but doesn't have the capacity to build stuff like this, my kids and I love the Korg Koassilator: https://www.korg.com/us/products/dj/kaossilator2/
You can usually get them on eBay for USD$60 - 70. You do need to bring your own speakers, but a pair of cheap PC speakers are good enough, and it's a good start on creating a whole synth + effects chain.
This one is a little more difficult to figure out for the kids as it is a "modifier" in the chain and they haven't quite wrapped their head around that concept yet. But still, it works great, has lots of features, and is really inexpensive for such a thing.
One of my favorite kids music toy is a mini piano where you have to replay one among 5-6 melodies (each melody is 5-10 notes that you have to memorize by ear — no lights involved) to unlock a little happy sound. This toy managed to keep both, dad and kid busy for a while.
So much text and not a single example, diagram, or demo.
I'm honestly skeptical this will work at all, the FOV of most webcams is so small that it can barely capture the shoulder of someone sitting beside me, let alone their eyes.
Then what you're basically looking for is callibration from the eye position / angle to the screen rectangle. You want to shoot a ray from each eye and see if they intersect with the laptop's screen.
This is challenging because most webcams are pretty low resolution, so each eyeball will probably be like ~20px. From these 20px, you need to estimate the eyeball->screen ray. And of course this varies with the screen size.
TLDR: Decent idea, but should've done some napkin math and or quick bounds checking first. Maybe a $5 privacy protector is better.
Here's an idea:
Maybe start by seeing if you can train a primary user gaze tracker first, how well you can get it with modeling and then calibration. Then once you've solved that problem, you can use that as your upper bound of expected performance, and transform the problem to detecting the gaze of people nearby instead of the primary user.
Sorry, I haven't gotten around to gathering examples yet. I ran the models on some example videos which is where the accuracy stats come.
Perhaps I have been jaded by the Mac webcam, I agree on most old webcams it wont be great but on newer webcams I have had success.
I did try a calibration approach but it's simply too fragile for in the wild deployment, calibration works great if you only care about one user but when you start looking at other people it doesn't work so well.
Good idea, it may be more fruitful to do that. At least then for the primary user we can be much more certain.
1. Punch cards -> Assembly languages
2. Assembly languages -> Compiled languages
3. Compiled languages -> Interpreted languages
4. Interpreted languages -> Agentic LLM prompting
I've tried the latest and greatest agentic CLI and toolings with the public SOTA models.
I think this is a productivity jump equivalent to maybe punch cards -> compiled languages, and that's it. Something like a 40% increase, but nowhere close to exponential.