you would think so! but that's only optimal if the model already has all the information in recent context to make an optimally-informed decision.
in practice, this is a neat context engineering trick, where the different LLM calls in the "courtroom" have different context and can contribute independent bits of reasoning to the overall "case"
Falconer | Lead AI engineer | Bay area, hybrid/onsite | $200-300k + equity
Falconer is building AI-powered knowledge tools for engineers. Knowledge management is a $30b industry, and we intend to build the tools that leverage AI to power the next generation of engineering knowledge. Our vision is to empower every engineer to write as well as the best technical writers, and enable every organization to have internal docs as great as the best external docs are.
We're currently a team of 8 from Stripe and Meta, currently turning our prototype into a generally available product.
Looking for an engineer experienced in evals, inference latency optimization, prompt caching, and LLM agent reliability. Or someone who just loves tinkering with the latest models and newly-published techniques, and wants to work somewhere you try them out in a real product that people use!
- Learn their default layout first, then iterate on your personal preferences. I similarly tried a bunch at first, but realized switching layouts constantly was making it impossible to build up the new muscle memory.
- Spend time practicing away from important coding tasks (e.g. gaming, writing, a side project, a one-off script, etc.). Being able to learn in a forgiving environment, outside of deadlines, gives you the space to allow mistakes and time to correct them
Falconer | Founding team (AI, infrastructure) or early-career engineers (any) | Bay area, hybrid/onsite | $125-250k
Falconer is building AI-powered knowledge tools for engineers. Knowledge management is a $30b industry, and we intend to build the tools that leverage AI to power the next generation of engineering knowledge. Our vision is to empower every engineer to write as well as the best technical writers, and enable every organization to have internal docs as great as the best external docs are.
We're currently a team of 5 from Stripe and Meta, prototyping the product with design partners, and hiring a few more engineers to round out the founding team. We've raised a $5m seed round and have several customers lined up, we just need to build it!
Falconer | Founding team (AI, infrastructure, full-stack, designer) | Bay area, hybrid/onsite | $150-200k
Falconer is building AI-powered knowledge tools for engineers. Knowledge management is a $30b industry, and we intend to build the tools that leverage AI to power the next generation of engineering knowledge. Our vision is to empower every engineer to write as well as the best technical writers, and enable every organization to have internal docs as great as the best external docs are.
We're currently a team of 4 from Stripe and Meta, prototyping the product with design partners, and hiring a few more engineers to round out the founding team. We've raised a $5m seed round and have several customers lined up, we just need to build it!
Falconer | Founding team (AI, infrastructure, full-stack, designer) | Bay area, hybrid/onsite | $150-200k
Falconer is building AI-powered knowledge tools for engineers. Knowledge management is a $30b industry, and we intend to build the tools that leverage AI to power the next generation of engineering knowledge. Our vision is to empower every engineer to write as well as the best technical writers, and enable every organization to have internal docs as great as the best external docs are.
We're currently a team of 3 from Stripe and Meta, prototyping the product with design partners, and hiring a few more engineers to round out the founding team. We've raised a $5m seed round and have several customers lined up, we just need to build it!
Falconer | Founding team (AI, infrastructure, full-stack, designer) | Bay area | Hybrid/onsite
Falconer is building AI-powered knowledge tools for engineers. Knowledge management is a $30b industry, and we intend to build the tools that leverage AI to power the next generation of engineering knowledge. Our vision is to empower every engineer to write as well as the best technical writers, and enable every organization to have internal docs as great as the best external docs are.
We're currently a team of 3 from Stripe and Meta, prototyping the product with design partners, and hiring a few more engineers to round out the founding team. We've raised a $5m seed round and have several customers lined up, we just need to build it!
What makes you say it's a "left-wing culture war item"? I've gotten the impression opposition to using AI-generated content like this typically comes from artists who might have been paid for similar content in the past
The calls per server is probably not the difficult part - this is the type of scale where you start hitting much, much harder problems, e.g.:
- Load balancing across regions [0] without significant latency overhead
- Service-to-service mesh/discovery that scales with less than O(# of servers)
- Reliable processing (retries without causing retry storms, optimistic hedging)
- Coordinating error handling and observability
All without the engineers actually writing the functions needing to know anything about the system (which requires airtight reliability, abstractions, observability).
I don't mean to comment on whether this is impressive or not, just pointing out that per-server throughput would never be the difficult part of reaching this scale.
[0] And apparently for this system, load balancing across time, which is at least a mildly interesting way of thinking about it
I'd be super interested to see an analysis using exactly the same methodology as this performed on various other cities (SF, NYC, LA, maybe some European cities like Amsterdam or Cologne) for comparison. I would bet Seattle is near the top for the US in walkability (probably behind NYC only)
There are some legitimate concerns about Warp throughout these comments (telemetry, business model, etc.).
But the one thing that really excites me is to have a full team working full-time on building the terminal that developers want to use. They're doing real user research, talking to developers, and taking feedback in forums like HN seriously - and using up millions of VC-dollars building a new version of this fundamentally important core utility. I'd much rather have that VC money go toward an attempt at a better terminal than some ML or web3 startup.
I think this doesn't usually happen? All the terminal emulators I've used usually open-source projects developed in someone's free time. Don't get me wrong, projects like Alacritty, urxvt, xterm, Terminator etc. are amazing for the funding they have (I think mostly $0?), but I'm super excited to see what a cohesive terminal based on real UX research can look like.
The interesting part about this idea to me is the scalability.
If you're someone looking to cold-email someone you admire, you're willing to spend a dollar to do that.
If you're a spammer who plays number games with your emails (maybe 0.01% result in a successful clickthrough), $1 per email is not even close to worth it.
I haven't thought through all the consequences yet, but this seems like a super interesting idea to me, with the potential to kill off the low-probability-of-success emails that I get hundreds of per day!
As a computer scientist who learned JavaScript this year, I strongly agree with the sentiment here. I feel like I often spend more time debugging module errors than my actual project.
Granted, once I understood the ecosystem more, I gradually began to grasp the reason these issues are around. But still, it does make it incredibly frustrating as a reasonably decent coder getting started in JavaScript.
AutoHotKey was my first programming language, years and years ago on my couple hundred dollar laptop. It wasn't much, but it was definitely a great introduction to basic CS concepts (if statements, loops, functions, etc.).
I remember using it to automate login for a variety of websites, and to automate resource collection for a game I played (Lord of Ultima, in case anyone remembers it).
you would think so! but that's only optimal if the model already has all the information in recent context to make an optimally-informed decision.
in practice, this is a neat context engineering trick, where the different LLM calls in the "courtroom" have different context and can contribute independent bits of reasoning to the overall "case"