skoocda

This misses one of the best mantis shrimp facts.

One of its receptors only detects circularly polarized light

But the only thing we know of, in the entire natural world, that emits circularly polarized light... is the reflection off the shell of the mantis shrimp.

Are there any resources you would recommend for learning about real implementations?

While the article is oriented towards defense planning, many of these points apply to any sufficiently complex software engineering project.

Generating several "competing constructs" used to be a wanton misappropriation of resources, but now it's not only viable... it's cheap. Comparing these constructs, however, has not necessarily become easier.

Leaders need to avoid using the base appearance of competence to accept any given plan (as they will all appear internally coherent), and instead spend more effort determining which difficult questions have been left unasked. As we learned from HHGttG, the answer is the easy part.

Not OP, but having the exact VM spec your agent runs on is useful for testing. I want to make sure my code works perfectly on any ephemeral environments an agent uses for tasks, because otherwise the agent might invent some sort of degenerate build and then review against that. Seen it happen many times on Codex web.

I have a degree in EE (2016) and am doing mostly ML engineering with a considerable amount of SWE tasks in my day-to-day.

Of my graduating class, very few are designing hardware. Most are writing code in one form or another. There were very few jobs available in EE that didn't underpay and lock you into an antiquated skillset, whether in renewables/MRI/nuclear/control etc.

We had enough exposure to emerging growth areas (computer vision, reinforcement learning, GPUs) to learn useful skills, and those all had free and open source systems to study after graduation, unlike chip design.

The company sponsoring this article is a contributor to that status quo. The complete lack of grassroots support for custom chips in North America, including a dearth of open source design tools or a community around them, has made it a complete non-starter for upskilling. Nobody graduates from an EE undergrad with real capability in the chip design field, so unless you did graduate studies, you probably just ended up learning more and more software skills.

But the relentless off-shoring of hardware manufacturing is likely the ultimate cause. These days, most interesting EE roles I see require fluency in Mandarin.

The one thing I really disagree with is the notion that there will be millions of identical AI images.

The next big step is continual learning, which enables long-term adaptive planning and "re-training" during deployment. AI with continual learning will have a larger portion of their physical deployment devoted to the unique memories they developed via individual experiences. The line between history/input context/training corpus will be blurred and deployed agents will go down long paths of self-differentiation via choosing what to train themselves on; eventually we'll end up with a diaspora of uniquely adapted agents.

Right now inference consists of one massive set of weights and biases duplicated for every consumer and a tiny unique memory file that gets loaded in as context to "remind" the AI of the experiences it had (or did it?) with this one user / deployment. Clearly, this is cheap and useful to scale up initially but nobody wants to spend the rest of their life with an agent that is just a commodity image.

In the future, I think we'll realize that adding more encyclopedic knowledge is not a net benefit for most common agents (but we will provide access to niche knowledge behind "domain-specific" gates, like an MoE model but possibly via MCP call), and instead allocate a lot more physical capacity to storing and processing individualized knowledge. Agents will slow down on becoming more book smart, but will become more street smart. Whether or not this "street smart" knowledge ever gets relayed back to a central corpora is probably mostly dependent on the incentives for the agent.

Certainly my biggest challenge after a year of developing an industrial R&D project with AI assistance is that it needs way, way more than 400k tokens of context to understand the project properly. The emerging knowledge graph tools are a step in the right direction, certainly, but they're not nearly integrated enough. From my perspective, we're facing a fundamental limitation: as long as we're on the Transformers architecture with O(n^2) attention scaling, I will never get a sufficiently contextualized model response. Period.

You might notice this yourself if you ask Claude 4.5 (knowledge cutoff Jan 2025) to ramp up on geopolitical topics over the past year. It is just not physically possible in 400k tokens. Architectures like Mamba or HOPE or Sutton's OAK may eventually fix this, and we'll see a long-term future resembling Excession; where individual agents develop in enormously different ways, even if they came from the same base image.

I've done a precursory skim of this and plan to start reading it in earnest next week. Looks comprehensive and accessible. Very excited.

e.g. HashCash

Regardless of net efficiency, that still entails collecting CO2 at a central facility (where it could have been dealt with in other ways, such as injection underground) and sprinkling it through the air as you fly over delicate ecosystems. I'm sure bankers see both as net zero, but condors might have more issues with your simpler workaround.

That gravimetric energy density is about 2 orders of magnitude higher than lithium ion batteries.

What exactly is your "social advertising bot"? Is this a bot that makes contextual comments like the one you're replying to?

Horses used to be smaller.

Przewalski's horse, the closest thing to a pre-domesticated horse, is about 660lb. A modern draft horse is over 2000lb.

50 mya, Eohippus was around 20lb - the size of a fox.

This was a phenomenal read. Thanks for linking

This is awesome! Presumably you can make this work with any interface that doesn't enforce the total internal reflectivity of a fiber optic cable, and therefore allows light to leak out. Instead of an air gap, have you tried experimenting with removing the cladding of the fiber optic cable, but keeping the core intact?

Alteratively, could you use a short segment of colored cladding that allows certain wavelengths to leak out more than others? I think that would allow you to encode each bend point as a different color-- which might require a different (more expensive) rx sensor, but could be useful for certain applications.

First few instances I tried are either returning no results, or only DDG results.

Error! Engines cannot retrieve results:

brave ( Suspended: too many requests )

google ( Suspended: too many requests )

qwant ( server API error )

If someone was running a supermarket and you came in day after day and never paid for anything, they'd get suspicious and probably follow you around all the time, too.

You don't pay for (almost) anything on the internet. Nobody does, because the requisite infrastructure doesn't even exist to allow you to pay for a page at a time. Until that changes, the creepy companies will keep finding ways to follow you around- both to earn money from advertising, and to ensure you're not abusing their systems.

Probably worth mentioning Deepmind's SIMA if you haven't seen it yet [0]. It definitely pushes the state of the art for agent behavior in a typical game environment.

Perhaps in the future we won't code NPCs any differently from player characters- they'll have access to the full action space, with long-term memory and natural language prompts guiding their behavior.

[0] https://deepmind.google/discover/blog/sima-generalist-ai-age...

Your github profile and website are awesome! Good luck with the job search

That's false. Every website needs a server to survive. Serving can be cheap, but it ain't free.

Yes, or use a ZK rollup between the publisher and the client to collect transactions off-chain and then settle when it reaches an appropriate amount.

Or even better, have a ZK rollup between each client and a single intermediary, who manages ZK rollups with all publishers. Then you, as a client, don't have to worry about large amounts being trapped in escrow, and publisher payouts happen more regularly.

We could call it "Mastercard"

Part of this issue is that most podcasts were designed to be free from the beginning, whereas music was always paid.

As a user, it makes sense to pay Spotify for music access, because Spotify, in turn, pays out a few fractions of a cent to the artist. But it doesn't really make sense to use Spotify premium for podcasts, when you can just download them directly or use various free RSS-based clients.

So why do so many people still consume podcasts on Spotify? 1. Because they're already on Spotify, so it's convenient, but more importantly... 2. Because Spotify realized they make more money if most of their users are paying, yet still consuming free (podcast) content. They don't pay out on podcasts per-stream like they do for a typical song. So ultimately, a premium user who consumes 100% podcast material is more profitable than a user who consumes 100% label-published music.

It's not a coincidence that Spotify has gradually taken over its homepage with podcast promotions, and promoted your "podcast library" as a first-class citizen, even placing it above your music library. It's all about profit margins, baby!

Any specific favorites? Or technical achievements that are worth calling out?

I feel like battlebit is laudable for getting 254 players in sync at a 60 tick rate

Indira Gandhi National Open University has over 4 million students, and in China about 40% of university students are in STEM. It's certainly not impossible that this is coming from a single institution.

Reach out to a monetization partner. They'll have dozens of ways to help find ad demand quickly. It will be more work than AdSense, but it will work- unlike AdSense. Here's a list of options: https://www.google.com/ads/publisher/partners/find-a-partner

Look for resellers or programmatic direct services to get started.

+1 for The Odyssey. I started it during an off year, almost as an intentionally-irrelevant book to pick up and forget about, but it ended up being a genuine gem; one I remember fondly, and often.

Try this tool:

https://www.movemap.io/explore/us

What made you drop zencastr?

GPT-3 and DALL-E. They're mind-blowingly great technologies that maybe aren't really "products" yet per se, but I'm truly excited about what they can do.

I think type 2 fun makes for the best stories. So, if you tell those stories and many people enjoy them, then from a utilitarian standpoint, it could certainly be a good idea!

I also tried and failed here. We ran our own speech engine with a custom model- but it's extremely expensive as a cloud service, and incredibly tough to reach acceptably high accuracy in different environments. Adding NLP on top of error-prone transcripts will multiply the error rate and lead to all sorts of weird actions.

I really think on-device models like we see in Android's Live Caption tool are a major privacy boon, and they're starting to reach an acceptable level of performance in Google's case. The main pathway to better performance is loading ever-more-massive models into memory, which isn't feasible for mobile devices but could be done on people's laptops in a meeting.