Agreed. One genuine question I have is: how is this not a five alarm fire at Google? What are the set of constraints that have made this unable to be solved?
The effort to fix it vs. the risk seems weirdly asymmetrical. Why do all the hard things right and then trip up at the last hurdle?
- Single point of billing and vendor is still king in enterprise - look at how much of Ant's revenue flows through AWS Marketplace
- With OSS, models may move to a more of an AWS-shaped problem: host with 5 9s, provide the infra and shovels, do it as cheap as possible, fit them neatly into rest of AWS stack. I think this is actually a huge risk to OpenAI and Anthropic, as more commodization gives hyperscalers more advantage
- Still owns a huge chunk of Anthropic and leverage over compute
With this in mind, I'm not sure why they need a frontier model vs. the playbook they did with Linux (i.e. Kimi-K3-AWS-Enterprise-Edition). They are kind of playing both sides right now I guess, but could really stir things up if they wanted to.
I think we'll see it more and more... open-ended Q&A / novel work is everyone's first experience of LLMs, but I don't think it's actually representative of most of the work people do. Coding kind of sits in the middle as it's novel but semi-structured.
Chat is great for novel open-ended work, but is terrible at concurrency - i.e. managing > 1 agent runs. As agents get cheaper, I suspect we'll see more more bounded processes where you want to run lots of items through them at once, and thus Jira-like things... (we ended up having this problem and building something like this, albeit more medieval themed [0])
We also found the other piece that Jira-style solves (and that chat makes super hard) is multi-player. I've yet to see a good implementation of a multi-player chat solution to agents working concurrently; and we found that an async model is the only sane solution to this.
I had to build a diffing engine for PowerPoint files in my first start-up. I initially tried to do this at the OOXML level, and one the weirdest things I found about the format is that the GUIDs are mutated by PowerPoint even if there have been no changes. The serialized OOXML is not really canonical and constantly changes as clients open and save it.
The simplest solution ended up being rendering images of the slides (not easy) and doing it with a slider.
I would love for someone to design a great case with a built in e-reader on the back. Most of the existing ones seem geared for displaying colour photos; I'd like one built for reading, powered by NFC. It should easily sync up with the phone itself to send over articles, books, etc. Bonus if I can write on it.
The effort to fix it vs. the risk seems weirdly asymmetrical. Why do all the hard things right and then trip up at the last hurdle?
reply