bbatha

Funnily sub-zero fridges are considered the samsung of the semi-commercial brands. They prioritize things like customizable doors and easy to break but consumer loved features like ice makers and water dispensers. You will never see sub-zero in actual commercial kitchens.

> say "new semantic information". It should be communicated, and you can't trust an LLM to choose the correct database, you need to specify that yourself.

But I can most of the time, and correct it if it chooses the wrong thing. That's the whole reason LLMs are faster. Its the reason we can give a paragraph prompt and get a kLOC PR back but only need to correct about 5% of it.

The value of writing isn’t always to communicate new information. It’s often to align everyone’s assumptions. For instance, when I say casually to a colleague or an agent “this change will require a db migration” they understand it’s to my teams primary application database. If I submit a design doc to a company wide review which database is changing is critical information. If the 300 bits were truly enough your agent or junior engineer would implement the wrong thing correctly as they often do.

Unified memory is the answer, it’s one of the cheapest ways to get >64GB of gpu addressable memory.

Don’t worry, vllm is also buggy on high end hardware

> The actual risk that the author does not even broach upon for investors... the thing that will actually torpedo this massive investment are the open source open weight chinese models that commoditize the entire endeavor.

OpenAI and Anthropic investors yes, however open weight models are good for cloud providers. They can turn the two large customers into direct ai services that can be spread across many customers and reduce the cloud providers overhead on ai services.

> future models can be trained to be even more aware of external knowledge

Then you need longer contexts, which is proving to a much more stubborn problem than general knowledge compression.

Exactly cd players in cars didn’t become common until the 2000s when anti skip became common place.

I’ve been asking for the volume feature since I switched from a blackberry in like 2009!

Resolve unfortunately only supports a few brands of raw. Om systems is not supported for instance.

You can. Davinci is more powerful for this too because you can copy the parts of the node tree. So you can have a base edit for exposure, white balance etc then apply your look as a copy paste on top of the clean part. In Lightroom and most photocentric workflows you get one prebaked pipeline.

It is. That was for the beta

For all the potshots about AI, this update is huge even if you take away the AI features. They basically added lightroom to this release. There's some polish before you'd want to change your subscription, but its really tempting. It may be the best photo management/editor on linux. Yes, I know about darktable and rawtherapee and I stand by what I said. They also added a ton of motion graphics stuff which from the beta seem to be enough to undercut a lot of basic uses of after effects out. The later two features are in the free release as well!

Imo vistas biggest problem was they underspecced the minimum system requirements. Most peoples experience with vista was on sub 1gb of ram machines with a single core, no gpu and 32 bits. Vista was an os for modern architectures first. Windows 7 had the luxury of those features being the default at launch.

What's time to first token? Raw throughput is usually not the problem in local setups in my experience.

Moreover the two lost shuttles included human lives. Better to blow stuff up with demo payloads now before sending up large contracted payloads or worse human beings!

Games will typically un at 4k or less which is about 8mp on the other hand it’s difficult to buy a stills camera with less than 20 mp, and >40mp is common. Most algorithms are n^2 in graphics as well so we wouldn’t expect a linear speed up. I’ve tried dxo, Lightroom, and topaz they all perform about the same so I don’t think it’s particularly unoptimized.

Interesting, IBM sales and consulting were infamous for sending in an army of blue suits in

I remain convinced we won’t look at project estimates as time based in software engineering as our primary cost estimate. And this is transition will happen rapidly. We’re going to shift to a capex/token spend model for project estimates where the business will say “ok I do want that feature for $1000 in tokens”.

A lot of enhancements came on the model side which in many ways enabled context engineering.

200k and now 1M contexts. Better context management was enabled by improvements in structured outputs/tool calling at the model level. Also reasoning models really upped the game “plan” mode wouldn’t work well without them.

I generally do this when I arrive at the agent getting stuck at a test loop or whatever after injecting some later requirement in and tweaking. Once I hit a decent place I have the agent summarize, discard the branch (it’s part of the context too!) and start with the new prompt

Newton knew a lot about investing for the time. He was a master of the mint for much of his adult life.

Its funny you mention excel, I see vibe coding in the business sense right now being a gateway to replace all of the ad hoc uses of excel. We've basically leveled up the quality of the software you can build before buying a SaaS product or a hiring an in house engineer.

I’m more interested in what happens when a language is designed specifically for llms? When doing vibe coding a lot of code is a lot more verbose than I’d do normally. Do we drop down the abstraction level because llms are just so good a churning out boilerplate?

Don't forget shortening copyrights would have a major impact on this issue as well! 25 year copyright would make most of these libraries public domain. Then the price would truly reflect what we would all like to be subsidizing: the new stuff.

Does the rights to Friends being traded around for the Nth time really benefit anyone?

> explain exactly how this doesn't end poorly

Explanations aren't sufficient either. The industry has burned that bridge. Strong contractual guarantees. Ceasing personal data collection operations, etc. etc. Concrete steps only. Thus far we have one concrete step that is proof of the opposite direction.

> f you need one in zig you can go that approach or use a tagged union just that zig is open about what an interface is.

Perl objects worked like this and the ecosystem is a mess as a result. There's a built in `bless` to add a vtable. There's some cool stuff you can do with that like blessing arrays to do array of structs vs struct of arrays and making it look like a regular object in user code. The problem is there are like 4 popular object libraries that provide base stuff like inheritance, meta objects, getters/setters etc and they're not all compatible and they have subtle tradeoffs that get magnified when you have to pull multiple in to get serious work done.

It’s several factors and all of your alternatives are true to some degree:

1. An h20 is about 1.5 generations behind Blackwell. This chip looks closer to about 2 generations behind top end Blackwell chips. So ~5ish years behind is not as impressive especially since EUV is likely going to be a major obstacle to catching up which China has no capacity for

2. Nvidia continues to dominate on the software side. Amd chips have been competitive on paper for a while and have had limited uptake. Now Chinese government mandates could obviously correct this after substantial investment in the software stack — but this is probably several years behind.

3. China has poured trillions of dollars into its academic system and graduates more than 3x the number of electrical engineers the US does. The US immigration system has also been training Chinese students but having a much more limited work visa program has transferred a lot of knowledge back without even touching IP issues

4. Of course ip theft covers some of it

I wrote a longer post a few months ago.[1] The tl;dr is a) computer aided design and manufacturing b) aspherical elements c) fluorite glass d) retro focus wide angle designs and e) improved coatings. Mirrorless lenses also beat slr lenses because they are much closer to the film plane — of course rangefinders and other classic designs never had this problem to begin with. 1: https://news.ycombinator.com/item?id=42962652

Edit: this is just for prosumer style cameras. If you look at phone sized optics that’s a whole other ballgame.

Depends on how you define quality. While medium and large format photography are extremely high resolution that’s not the only factor. Space age lenses were significantly lower resolution than the film. Modern mirrorless lenses are starting to come close to being able to out resolve film but still aren’t there. Meaning that you get more functional resolution out of modern digital. Digital also beats the pants off film for dynamic range and low light. That said the noise (grain) and dynamic range fall off in film are more pleasing than digital to most eyes. So it’s not all about technical specs.