I felt taken by the change in the system model personality and writing style compared to opus, but I also found it to be much less impressive than I was expecting - let alone that the cost was incredibly high when not given for free.
Are you sure your reaction is not primarily to the improved ergonomics of Fable?
We're building the AI agents that run insurance - a $1T+ industry still run on PDFs, email, and manual review. Our agents read the documents, pull from the systems underwriters already use, and actually complete the work. In production at top-10 carriers and F100 brokers; ~10x growth in the last year; backed by a16z and YC.
The hard problems: keeping an agent reliable across dozens of tool calls; orchestration + evals + reliability on top of frontier models.
We've rebuilt the product from scratch more than once as the models got better!
I’m an ML Engineer with a software engineering background and an MS in EECS. My recent work has been around deep learning, computer vision and data workflows for real-world ML systems. Before moving into ML, I worked as a backend Software Engineer building payment processing systems with Java/Spring, REST APIs, and SQL databases.
I’m looking for Applied AI/ML Engineer, Computer Vision Engineer, or SWE roles. I’m especially interested in teams building practical ML systems, scalable inference/data pipelines, backend platforms, or applied AI products.
Good game, but the fact that there is a target that you have to achieve in a fixed number of rounds changes the math.
For instance, if all I care about is satisfying your win conditions, and I am above $20,000, then I should just pass till the end. Similarly, if I have been unlucky (even with Kelly bets) for the first 14 rounds, I should go all in on the 15th round.
People have tried all kinds of techniques for VLIW, including techniques that are much better than a purpose built AI, AI isn't a magic silver bullet. Fundamentally there's no reason you can't analyse a piece of code to death, and maximally extract parallelism out of it
The fundamental issue is that there simply doesn't exist enough information to be able to extract the necessary parallelism without a rewrite, its the same issue as trying to autovectorise. You can do it to some degree, but it doesn't work in practice to be able to fill out a very wide architecture with reasonable efficacy
The SIMT programming model has proven to be much more successful vs trying to autovectorise or mash things into a VLIW architecture
The same process we've used for centuries? You got something better? I'm sure you don't. You are also ignoring that the 12 random schmoes are selected by both sides attorneys and given instructions and facts and can still be overruled by a judge (under circumstances) plus the possibility of appeal.
A more practical question: has anyone used this in production in internal apps at work? UIs like this look nice and have a flare that I like, but I guess it feels weird if I want to introduce this with my team over something like tailwindcss and rolling our own components
Kind, smart, low-drama people, seeking the same to help make writing tools authors dream about.
Our users have published thousands of books using Sudowrite, and we're building a sustainable company, not chasing unrealistic growth targets set by VCs.
* We help the next generation of storytellers tell better stories.
* We believe the future of writing is AI & human collaboration.
* Co-founders both had prior exits.
—————————————————————
Hiring: Senior Full-Stack Developer & Mobile Developer
* Build imaginative tools that help writers turn ideas into stories.
* Work across the stack: from expressive, responsive interfaces and text editing to the systems and AI infrastructure behind them.
* You should have: Great taste and product sense, lots of agency, strong React and backend experience, and a sense of humor, adventure, and hope.
I think it's pretty obvious from the context of "if you had technical gaps (say, you couldn’t write CSS)" that this isn't about someone that regularly works on frontend web apps, it's about someone that is having to do something that is not normal for them. E.g. I mainly work on HPC code, maybe once in a year or two I need to throw together an animated poster to accompany a printed one. It's obviously ridiculous to claim that this means I have to learn frontend web dev.
Backend engineer with 18+ years of experience going deep on complex systems to find the highest-leverage improvements. Led greenfield projects from design through deployment at fintechs you've heard of; built payment and accounting systems processing billions of dollars. Wrote the book on integrating Stripe with Rails.
Generative AI is.. not better than the best humans at coding even today, let alone "for years" - I'm not sure how to even have a conversation if you believe this. Better than the worst humans, sure. Better than p50, probably also true.
> How does a library writer testing that their library works with a specific version of a compiler that was already released have anything to do with a compiler implementation providing special treatment for that library?
Not special treatment for the library, but the library being able to assume that specific UB handling in certain existing compiler versions exists, can be relied on, and is compatible with the library's assumptions of such behavior, as long as those are the only versions they officially "support."
So I feel like I need to write a blog post about this, but succinctly I think there's a good argument to be made that the issue wasn't the compiler despite popular wisdom. I don't even think it was the nature of unpredictable memory access times either as the itanium has a ton of special architectural hardware to handle unpredictable memory accesses (a lot of which are essentially some of the primitives that an OoO core uses for internal bookkeeping, just exposed architecturally).
I just think the arch has a similarity to archs like cell where it was planned for a world without the end of dennard scaling and just stopped making sense when we weren't targeting scaling to 10Ghz consumer CPUs and beyond.
The relatively fixed clock period that makes sense post ~2006 also means that the CPU architecture of that made the most sense ~2006 (Tomasulo OoO cores) continues to make sense, with most of the process gains going to just making bigger, wider cores.
What is the typical job title and/or skillset for this?
It can be any one of many jobs depending on how high close to the metal one's focus is, but the highest headcount role is usually SRE/Infra/Ops with GPU knowledge sprinkled on top. That is to say Linux sysadmin, networking, fleet management, scaling, incident troubleshooting, etc.
Please stop using HN for self-promotion. It's specifically against the guidelines and we have to ban all accounts and domains that engage in it, which we've done now. https://news.ycombinator.com/newsguidelines.html
Itanium never made it into the consumer-class hardware that was XP’s audience. AFAIK, Intel never even published a roadmap for that to happen!
Maybe a proof of concept they shipped as a demonstration of loyalty to Intel?!