zild3d

The rate of improvement of all models is converging to zero.

Curious where you draw this conclusion from? Most benchmarks still show continual steady progress https://metr.org/time-horizons/

Imagine if you were given a layer of abstraction that randomly flipped bits from time to time... That's not an abstraction, it's random programming.

Computing has never really been deterministic all the way down. Storage gets corrupted, RAM has soft errors, networks drop/reorder packets, schedulers race, caches go stale, distributed systems partition, query planners change plans, ...

Obviously those become tolerable because we've developed layers of contracts and understanding the bounds around them. Error correcting codes, checksums, retries, consensus, idempotency, false-positive rates, SLAs, etc.

If the abstraction is “delegate this task to a junior engineer, analyst, lawyer, designer, or support rep,” you dont expect deterministic behavior. There's review, constraints, escalation, checklists, tests, and accountability.

defence - why would anyone buy US weapons after Greenland and Canada

Huh? US foreign military sales are up at all time highs

"Total exports by the United States, the world’s largest supplier of arms, increased by 27 per cent. This included a 217 per cent increase in US arms exports to Europe, according to new data published today by the Stockholm International Peace Research Institute (SIPRI)"

[https://www.sipri.org/media/press-release/2026/global-arms-f...]

Iran's 10-point plan (that no one else has agreed to)

whats surprising about that? most of the minor version updates from all the labs are post training updates / not changing knowledge cutoff

to be fair this is a pretty hot topic, that the work is shifting to the prompt/agent thread but then thrown out / not captured

greg brockman https://x.com/gdb/status/2019566641491963946

6. Work on basic infra [...] there's a lot of infrastructure that currently go around the tools, such as observability, tracking not just the committed code but the agent trajectories that led to them, and central management of the tools that agents are able to use.

peter steinberger (clawdbot/openclaw) https://newsletter.pragmaticengineer.com/p/the-creator-of-cl...

Peter now views PRs as “prompt requests” and is more interested in seeing the prompts that generated code than the code itself

economically disadvantaged countries, enthusiastic consumption of the local cuisine, and a subsequent return home marked by self-congratulatory reflections on how much they have supposedly "learned" about other cultures.

There can certainly be a quite shallow "instagram" quality to some traveler's trips, but it's also clear an economically disadvantaged country benefits mutually from this, and if it wasn't they'd be restricting tourist visas, etc

it's the best ____ we've ever made

POST /superuser/admin?permissions=all&owner=true&restrictions=none&returnerror=no

cool thanks, we've been considering dynamo global tables for the same. We have S3 replication setup for cold storage data. For primary/hot DB there doesn't seem to be many other options for doing local writes

active/active? curious what the data stack looks like as that tends to be the hard part

the downstream effect of that is less startups

These smaller models are great, but they really feel like talking to a toddler sometimes!

You're using the toddler and the model wrong. I love talking to my toddler, probably more valuable conversations than I've had with any other person. But it's not the same use case as asking a professor a question in their field

| Gemma 3 270M embodies this "right tool for the job" philosophy. It's a high-quality foundation model that follows instructions well out of the box, and its true power is unlocked through fine-tuning. Once specialized, it can execute tasks like text classification and data extraction with remarkable accuracy, speed, and cost-effectiveness.

huh? I'm saying adding rules and regulations reduces competition yes, by definition it adds barriers to entry. We can argue how high those barriers ought to be.

Highly regulated industries: healthcare, banking, aviation

Less regulated industries: web software, e-commerce, entertainment

It is easier for startups to get started in the latter, harder in the former.

What do you expect from an unregulated capitalistic system.

Competition, fortunately

I see it more like the goal is to build cool stuff, learning electronics is the happy unintentional side effect as you're exposed to concepts relevant to what you want to do.

e.g. I want to build a cool robot with my kid -> oh why can't you just wire the motors directly to arduino output pins -> oh motors need a lot of current to run ...

(btw have never heard voltage be called tension, TIL)

I think it's actually a fantastic intro to electronics. There's nothing you can really do with "just arduino programming", the whole point is it lets you interface with the real world and therefor encounter electronics problems by default

The article even touches on that in the first hello world

This simple exercise it by itself incredibly interesting that opened a series of questions:

Q: Why is a resistor needed? A: High current and increased temperature damage its delicate heterojunction structures, which eventually cause it to burnout Q: What happens if the polarity is inverted? A: Similar to a normal diode, current will not flow and the LED will not light up. As long as this reverse power is not high, the LED will not burn and can still be used with correct polarity afterwards Q: How to interpret its data sheet? A: There are several interesting aspects its datasheet, like the LED’s wavelength curve, operating current and voltage, etc

The puzzle is why

The feedback loop on novel/genuine breakthroughs is too long and the training data is too small.

Another reason is that there's plenty of incentive to go after the majority of the economy which relies on routine knowledge and maybe judgement, a narrow slice actually requires novel/genuine breakthroughs.

Sounds like it'd be pretty easy to vibe-code the site you want

really great, would make for a great tower defense style game as well. Start with few resources and learn what each capability can do. Defend against more complex/advanced threats over time.

Is the equipment efficiency meant to capture e.g. using a $1M missile to shoot down a $1k uav/rocket

isn't this essentially https://openrouter.ai/

the US. Getting dragged into an Israeli war

A lot of people saying this, what would this actually entail? My money is much more on this being a "1 and done" exchange. Iran poses very little threat now, launchers being taken out everyday, leadership chain wiped out, seemingly no other Iran allies getting pulled into the fold

I want an AI that I have similar opinions to, which is obviously tough. It's like working with someone on their first day.

Most of what you're describing does apply to humans on the first day, and ais on their first day. If you aren't capturing these preferences somewhere and giving it to either a human or the ai, then why would they somehow know your preferences? For ai, the standard thats forming is you create some markdown file(s) with these so they only need to be explained once, and auto provided as context.

Rockset - https://openai.com/index/openai-acquires-rockset/

send a map pin :P

Not op but I would also just use an ESP32 since they're super cheap, I already know how to work with them, and there's a pile of em in the drawer.

If producing multiple, yeah esp32 is overkill, but when cost is <$5 wouldn't be worth the time to make the project harder to get started

well for now its either small device that uses APIs or Paddington Bear needs a backpack for his GPU

the need for translating vague desires from laypeople into a precise specification will remain—there's nothing on the AI horizon that will solve for this

LLMs are very good at exactly that. What they aren't good at (I'll add the yet as an opinion) is larger systems thinking, having the context of multiple teams, multiple systems, infra, business priorities, security, etc.

He mentioned a rough estimate of one per century per galaxy. Estimate for average stars per galaxy is 100 million, which would be 1 in 10 billion stars every year

Are you trying to tell me this Rubbery Lasagne isn't a real recipe?

https://www.recipeninja.ai/recipe/r_LZarKW1PMNlSlx/rubbery-l...

All Ingredients

- rubber cement, 2 cups

- water, 1 cup

- lasagne noodles, 1 box

- shredded mozzarella, 2 cups

...

Step 2

Prepare the rubber sauce by mixing 2 cups of rubber cement with 1 cup of water in a saucepan over low heat until thickened.