davedx

Lots of RLHF - reinforcement learning from human feedback.

Also: "distillation" - seeding or running training sessions on the output of frontier models.

Very nice. Do you have a page comparing it to pi?

Think this is missing a URL?

There's a lot going on in energy, manufacturing, mining, and of course defence at the moment. For example, we're still making progress towards economically positive fusion power; Vehicle-to-Grid for EVs is starting to roll out as trials; hydrogen is still being seriously looked at in different parts of the energy supply chain; the first flight of a mid-sized electric aircraft happened the other day; a "quantum" based navigation system that doesn't need GPS did a couple of trials in the last weeks; laser based mining, new geothermal tech, and lots and lots of autonomous/drone defence tech is being developed.

Then there's also "newspace", an interesting area to watch!

Nope. The projects I'm on where we use it, we're carefully migrating to the newer models. Where we can we test with evals to try and get an understanding of how the models have changed.

It's not all roses -- I've seen some regressions -- but generally the 3.x Flash models are pretty great for our use cases.

The great thing about LLMs though is it's incredibly easy to diversify and have fallbacks. But of course that means additional costs, mostly centered around engineering efforts to test and integrate them.

https://octoloops.com - AI outbound and community outreach for indie/vibe coders. Unfortunately it's not going well: people sign up, but then they don't use the app. I'm running out of ideas to try quite frankly.

Don't make me laugh. Who has actually sat in Congress and told congress leaders to push regulatory capture?

Not Mistral

Mistral is an interesting AI company because they clearly have a contrarian business strategy to the other AI labs. They're also landing big customers in Europe for the right reasons. People dump on them because they're not benchmaxxxing which is pretty shortsighted - do you really want to be in a benchmark arms race with China, or do you want to make money and deploy sovereign AI compute in Europe?

I draw a man in China in 1819, and it said he "died of the war in Oman" without any explanation of how that could have happened. What a load of crap

There are some subtleties though. For example, with Gemini search grounding (this was a few weeks ago), you cannot give it a domain whitelist, only an excludelist -- with Anthropic's API you can do both. For somewhat niche search tasks like "Find LinkedIn profiles matching this ICP", Anthropic wins there.

That being said, Anthropic is also so insanely expensive for everything I ended up switching that particular part to Exo instead...

It's just a shame we still don't have the computer for the PITS, isn't it?

I guess the closest thing we have is an iPad - my mother can just about use tablets, but I have to say onboarding onto a tablet these days is full of gotchas and fraught with peril too if you don't understand computing.

I'm reminded of "The Dream Machine" by Licklider - if we read that today, we'd get this odd sense of having come very close, but yet missing by miles, many of the goals set back then.

Anyway, yes, monospace typesetting...

Yeah. The problem is, they have 7 cross platform UI frameworks...

When I rejoined X at the start of this year I consciously decided not to install the app on my phone. It's addicting enough in the web browser on my laptop.

So I have zero social media networks on my phone and, looking at everyone around me who does, I'm very pleased with this decision.

"Today, we outline our early research in intrinsic discovery, an approach for models to autonomously explore an environment and generate their own training experience. We use reinforcement learning to continually push a model toward novel states, then use the resulting experience to train a world model."

The company’s quantum navigation system maintained one nautical mile positioning accuracy over the mission duration, surpassing navigation-grade GNSS backups by more than 10 times.

This part seems relevant - they make explicit the difference between their HARNESS.md document and AGENTS.md. It's actually interesting: https://habitat-thinking.github.io/ai-literacy-superpowers/p...

This is a great idea. Anything that makes this easier is a win for ops teams.

I hang out on various entrepreneurial Reddit subreddits, both to get an idea for what people are building, problems they are having, and occasionally post about my own pain; and also to promote my own product. Every now and then I'll lurk in subreddits that are just stuff I'm interested in too.

I've also gone back to X/Twitter quite heavily (but trying to limit usage and avoid posting in anything politics related). It's hard not to get sucked into it.

Then occasionally sites like the Guardian for general news.

I'm sorry but that's a bit ridiculous. I'm still using an M1 Pro with 16GB of RAM and I have quite heavy usage - docker, other heavy native apps, a bunch of VS.Codes and coding agents, things running in the terminal, background processes. It's just a big exaggeration that general regular productivity apps "easily gobble up over 16GB"

(Edit: saying this as someone who just paid for a new 128GB M5 Pro Studio -- but that's for LLM's, not regular work)

Yeah, I was about to say "or an iPad Pro", but the Neo is shockingly cheaper than an iPad Air.

There's definitely an option for everyone if you want to go down the "fixed supercomputer, remote terminal" route.

I'm actually working on this problem by building OctoLoops (https://octoloops.com) to help find channels to promote the app and get users. It's working so far (the main agent I'm using is communities, which finds threads on Reddit and Indiehackers to post comments on). It's all human-in-the-loop to avoid being spammy and try to be genuinely helpful to people instead while also promoting the tool.

I'd love to hear your ideas on whether it would be useful for you too. Let me know if you give it a try (email's in HN profile), and good luck with your service!

I'm iterating hard on OctoLoops (https://octoloops.com). 30+ signups, and a handful of users are active and replied to my outreach positively. It's encouraging and I dogfood it but still very early days.

The core idea is building is now easy so let's help people with the hard part: getting (paying) users! It's a marketing harness where you "loop" marketing activities like posting to online communities or direct outreach. Human in the loop so attempting to balance the usual tightrope of reasonableness and being spammy.

Feedback very welcome.

(Never saw myself working on martech but, here I am)

Nope. I occasionally do human review of my own code and very occasionally of other peoples', but for the most part it's all 100% automated. I use Codex's Github code review integration and it does a pretty decent job of spotting P1's and P2's.

I have subscriptions for ChatGPT Plus and Claude Pro, so about $40/month. I use them for coding mostly, switching between them frequently and heavily. On the occasional times I run out of tokens I'll go outside and take a walk or spend time with family.

Been doing it this way for ~the last 6 months or so.

I do also use the API's for GPT and Gemini for some side projects, but the spend there is highly variable. It rarely exceeds $100 in a given month.

"I have so many half-baked ideas for tools or services I barely need, sparked by a combo of unaddressed workaholism and techno-optimist wishful thinking."

This part resonates. AI has been such an accelerator that just yesterday I told Claude to make a brand new "Christian male meal planner with grace quotes from the Bible", bought a domain name, had the whole thing ready to go then thought "wtf am I doing, I already have half a dozen other side projects in the go, this is insane!" and just stopped it.

Human focus is a precious, finite resource. The productivity boom enabled by AI tools is probably going to benefit those capable of laser-like focus the most. Scattered thinkers (like myself, or perhaps the original article author) are in for a struggle at times.

People and pundits have been dooming and bearing on Nvidia for as long as it's been around. It increased in intensity when gpus were used for large scale crypto mining and it became material to their operations, and continued as AI ("the bubble") started to really take off.

Over those years, my NVDA stock has been by far my biggest winner. I'm now up more than 1500% on it.

Let the dooming continue

I'm working on a bunch of stuff, but most interesting is maybe https://Octoloops.com - it's a sort of marketing harness for indie devs/SMBs. Still early days yet, feedback very welcome!

"~4× faster (up to 8× in specific workloads) than Spark and 94% cheaper on infrastructure costs"

No way in hell are the majority of Claude Code users burning 10k worth of credits. Many of them probably barely use it. There'll be a bell curve, and we have no idea what it looks like.

Come on. That doesn't go anywhere near explaining a million and a half LOC