Richard Bishop

I'm on Bluesky now: bsky.app/profile/richar…

What I love about this rage bait, which happens at least monthly these days, is how much of an absolute self-own it is by the author.

@JaredSleeper As a technologist I believe open weights models will win in the long run and both compute and models will become commodities. Though I acknowledge markets have the much harder task of figuring out how long proprietary labs will reign for and making money on that.

@JaredSleeper I agree with that. I'm quite smitten with the open weights models but acknowledge it's also a lot of work to optimize because the development of inference engines move at breakneck speeds. It only becomes a noteworthy trend if the recent behavior of frontier labs irk enterprises.

Switching to open source models is the new outsourcing to India.

@dampedspring The underlying themes that enabled a significant chunk of your list: - Moore's law never ended despite everyone declaring it did (it was Dennard scaling that ended) - NAND Flash memory became ubiquitous starting in mid 2000s

What were big aha's that changed your view of the next 10 years over your lifetime? And when I say aha immediately allowed you to picture 10 year from now Some of mine 1. Beta of Netscape web browser 1995 iirc 2. DSL vs dial up, always on internet 3. Amazon 4. Napster 5. AOL for my kids 6. BlackBerry 7. 9/11 8. Fiber speed 9. HFT/Rentech, Hull, Knight 10. iPhone 4 and smart phone Internet 11. QE Era and the permanent priority of financial stability over sound money 12. Palantir at BW in 2011 13. Big Data/web scraping /compute/ml offsite with Brevan Howard in Lake Como in 2016 which gave me AI 14. ChatGPT moment in winter of 2022-2023 Those are the ones that impacted my future outlook. I imagine CT followers will find it funny I don't include BTC, Blockchain, and Crypto but while I get it. I've never had an AHA moment like these above. Still not sure it matters at all. But if it does my aha moment was spring of 2017 when the guy who owned my gym told me about his bitcoin mining activity. Since the. Bitcoin has rallied 16% per annum with stupid drawdowns and low ratio. But again. Not the point of this little review. So relax CT guys. I didnt miss it I just don't care as much as you.

@dampedspring Also I really enjoy reading your market thinking, but don't follow because if I do my algorithm gets dominated with finance noise.

Loaded question. I just heard about the idea of premium tokens and it reminded me yet again how early we still are in AI. For that reason I picked zero. Frontier Labs controlling token inputs (context + apps that generate it), the function (models), and token outputs in the long run is wishful thinking. It's a misalignment of incentives and entirely confusing. It would be like if clouds billed in CPU cycles and cache lines. Outcomes are what businesses want. A healthy, layered ecosystem will eventually emerge. Consumers of the upper most application layers won't think in tokens at all. Right now the Frontier Labs are being greedy in wanting the whole cake. Software/algorithm improvements will also drive efficiency improvements many orders of magnitude that increase the useful lifespan of current hardware. A new, but far less excruciating problem of picking the right model for the task will emerge. That's setting aside what the hell is a premium token even. Tokens are nothing but fixed width data.

What's the fair price for an AI "token" given its utility measured as ROI

@brian_armstrong /join x17 1v1 msg me Casters turn gaming into a non-zero sum game for the fans at least. The SC communities are very lucky to have guys like Artosis, PiG, and Winter.

In my teens and 20's I would spend way too much time playing Starcraft and Civilization. Harvesting resources, building things, and expanding was super addictive to my brain - to an almost unhealthy degree. Later I realized that entrepreneurship and business is the ultimate game. It scratches the same itch for me (resources, building, expanding), but you're actually contributing to humanity at the end of the day, which can be much more fulfilling. Business is also much more positive sum than video games. In Starcraft, the other player has to lose for you to win. In business, there is competition, but in a growing market there can be multiple winners. And gains compound long term (it's a infinite game) instead of starting over each time. Now days I prefer to watch pros play video games to unwind, instead of playing video games myself. But a quick game can still be fun here and there to unwind. By contrast, the game of business is played over many decades.

@realcalebwin Worst case scenario is that we cede ability to create new abstractions to only frontier labs.

That depends on how much we help/let the big labs abuse their position. If fine tuning and RL become ubiquitous then what's under the hood can continue to evolve. Otherwise there's only room to evolve abstractions during context management and inference and the foundations ossify.

The problem with file systems being the abstraction for Agent sandboxes is, well, they require files. Files are the string that when pulled untangles the whole sweater: here comes a shell, POSIX I/O, hard coded linker paths in your builds, and the worst of all: package managers. LLMs of course can sort through all of this for you but why waste time and tokens on it? These superficial sandbox startup benchmarks mean nothing when an LLM spends minutes reaching around in the dark. I understand why sandbox startups choose the file system as the abstraction: it's ubiquitous and is a large tent for a business looking for traction. In the long run there's far more to gain by taking a stance and building an opinionated abstraction. The Overton Window for dev tooling has never been more wide. Perhaps it is time to seize that rather than burden ourselves with Dockerfiles, package managers, and all the rest of the baggage of yesterday.

@jhleath It's true. You're in too deep now.

@bitsandhops listen, everything about building this has been bad since I started working on POSIX stuff in 2016

we've just added a slew of software that you can now use out of the box on Archil's serverless execution. this means that you can use git, docker, compile and run rust/go/etc. what's missing?

@_j3ssica12 @awscloud I don't. A lot of it likely comes from having worked on successful systems, both commercially and technically, over long periods of time and dealing with the tensions that arrive from that.

@bitsandhops @awscloud A bit of a weird question: do you know of any books that explore more in depth this topic of asking good questions regarding architecture?

During my time at @awscloud I tried to be an astute observer of how the Senior Principal and Distinguished Engineers carried themselves in new projects/teams and design reviews. This has made a lasting impact on how I operate as well as evaluate others. All of these "Very Senior" engineers are smart but the absolute best ones were humble, secure, incredibly curious, and appreciated the history and path of a system. They can jump into whole new domains and immediately win over the people under them that ultimately do the work. The worst Very Senior engineers show up and inflict their insecurity by asking divisive and superficial questions like "why didn't you use {{language foo}} or {{database bar}} or {{platform qux}}?" These very questions reveal their inner fear: they don't have the depth and breadth their position requires.

@thesammiller @awscloud There's a baz in between bar and qux that I accidentally skipped.

@bitsandhops @awscloud Today I learned that foo and bar have a third called qux

@Carpediemwonder @awscloud This is hard to answer because success is very personal. I also haven’t worked there in almost 5 years and I’m sure the culture has changed, but a willingness to tackle unglamorous work has always served me well.

@bitsandhops @awscloud Hey Richard, If I were to work at Amazon. As a person what qualities should one possess to be successful in the role ? Not tech stuff but like individually.

@ranman @awscloud Certainly for the ones that have spent their entire careers there.

@bitsandhops @awscloud The best ones also understand the business model and not just the engineering decisions. I fear too many of the senior engineers and engineering leaders have only ever worked at Amazon though. AWS should have a program where the best PEs go work for customers for a year or two.

@viggy28 Why not SlopOps

DevOps. FinOps. RevOps. You know what’s next. TokenOps.

@mjackson @isaacinthesky Go’s socket types use much more memory and copy more often than an optimal Zig/Rust/C/C++/D/Odin implementation would.

@isaacinthesky Go was literally made for this kind of thing. You think a Zig/Rust rewrite would be an improvement?

i do take anurag's point here, and it makes sense, but a _load balancer_ is among the strongest possible places to think about a rewrite in in zig/rust

@waymodriver @awscloud Oh gosh do I know that feeling and it hurts in a good way. I can recall a time Matt Garman did this to me back when he was still leading EC2. A good reminder that we often needlessly create our own local maxima and minima.

I was also at @awscloud and was very intentional about observing ultra successful (L8+) engineers and leaders. I agree with what you said. I also think there is some “je ne sais quoi” element to it. Could never really understand how they are able to ask some extremely simple yet important questions that I being closer to the problem for months completely missed