HNHacker News
TopNewBestAskShowJobs

benzguo

481 karma · joined February 18, 2013

https://hraness.com
submissionscomments
benzguo··on Velocity Coding
There's a new kind of coding I call "velocity coding", where you fully give into the computer, embrace exponentials, and forget that the ego even exists. It's possible because the LLMs are getting too good, but also, importantly, YOU had already gotten too good at software engineering. Before you were just a bored apathetic engineer at a big corporation, but now you've quit your job and you know you're going to raise an absolutely monster seed round, because you now have the productivity of 10 S-tier engineers from Eastern Europe, and you feel like a god. Your mind races constantly with tasks, which are now just prompts (everything is just a prompt, really) and your waking life has become so simple: you are always stream-of-consciousness typing into cursor, or claude code, or slack, or linear, it doesn't matter where, because you know it will end up in the cosmic assembly line of pure creation, for which you are just a vessel – a single token in the vast computation stream of the universe...
benzguo··on Ask HN: Freelancer? Seeking freelancer? (June 2025)
SEEKING FREELANCER | Substrate Labs | NYC or Remote | https://www.notion.so/Substrate-Labs-Founding-Engineer-2067b...

Founders Ben [0] and Rob [1] have been friends since meeting at Venmo in 2013. Ben worked at Stripe for 8 years. Rob worked at Fin (human-in-the-loop AI executive assistant, before LLMs) and was founding engineer at Substack. We started our first product, Substrate in 2023, a platform for building agentic workflows, and raised an $8M seed round led by Lightspeed [2]. We were pushing the envelope on both infra (serving multi-model workloads efficiently) and developer experience (ahead of our time in several ways) [3]. Though we found solid traction with startups, we realized inference was a red ocean especially upmarket, and decided to spin down the product/team at the end of 2024. We've spent the last 6 months prototyping and discarding ideas, and we’re now gearing up to launch our latest vision, a consumer AI application.

[0] https://www.linkedin.com/in/bgdotjpg [1] https://www.linkedin.com/in/robertkcheung [2] https://lsvp.com/stories/substrate-building-compound-ai-syst... [3] https://docs.substrate.run/

Email me (ben [at] substrate) with your resume.

benzguo··on Ask HN: Who is hiring? (June 2025)
Substrate Labs | Founding Engineer | NYC or remote | https://www.notion.so/Substrate-Labs-Founding-Engineer-2067b...

Founders Ben [0] and Rob [1] have been friends since meeting at Venmo in 2013. Ben worked at Stripe for 8 years. Rob worked at Fin (human-in-the-loop AI executive assistant, before LLMs) and was founding engineer at Substack. We started our first product, Substrate in 2023, a platform for building agentic workflows, and raised an $8M seed round led by Lightspeed [2]. We were pushing the envelope on both infra (serving multi-model workloads efficiently) and developer experience (ahead of our time in several ways) [3]. Though we found solid traction with startups, we realized inference was a red ocean especially upmarket, and decided to spin down the product/team at the end of 2024. We've spent the last 6 months prototyping and discarding ideas, and we’re now gearing up to launch our latest vision, a consumer AI application.

[0] https://www.linkedin.com/in/bgdotjpg [1] https://www.linkedin.com/in/robertkcheung [2] https://lsvp.com/stories/substrate-building-compound-ai-syst... [3] https://docs.substrate.run/

Email me (ben [at] substrate) with your resume. Competitive comp with meaningful equity.

benzguo··on Show HN: Steel.dev – An open-source browser API for AI agents and apps
beautiful docs + api ref! what are you using? (cool that you're doing open-source browserbase also, excited to check this out)
benzguo··on The Cognitive Overhead of LangChain
I tried LangChain for the first time. For a simple example chaining two LLMs, it takes 60% less code and runs faster on https://substrate.run
benzguo··on Launch HN: Undermind (YC S24) – AI agent for discovering scientific papers
This is really cool! Both of my parents are cell biologists, and I've done some time in labs as well, so a lot of paper exploring and reading in the family. "Review" articles are a good index but something more on-demand makes a lot of sense, I can definitely see this being extremely useful.
benzguo··on Show HN: Animated captions for any audio file
You've probably seen those animated captions underneath videos all over the place. This is a super minimal demo of that, also with chapter summaries. Just one file hosted on val.town! https://www.val.town/v/substrate/subaudio
benzguo··on Generative AI Spanish Lesson in 70 lines of code
Hi! I put together this little app using Substrate – https://substrate.run – and Val.Town to help me brush up on my Spanish vocab. It's pretty neat to be able to generate text, images, and audio, and then ask an LLM to generate a website. Fun alternative to building out the UI yourself – ask an LLM to do it! I do this all the time now, asking the LLM to stream its output as Markdown or HTML.
benzguo··on Orch: a Rust framework for LLM orchestration
Neat! Always love seeing new takes on this, in new languages.

Check out Substrate - it's an orchestration framework that also runs the inference. https://substrate.run

benzguo··on Show HN: Semantically map JSON objects using an LLM
This was inspired by another HN post: https://news.ycombinator.com/item?id=40972130

I thought it'd be interesting to try implementing the idea with just an LLM call.

benzguo··on Shapeshift: Semantically map JSON objects using key-level vector embeddings
Put together a quick version with an LLM, using Substrate: https://www.val.town/v/substrate/shapeshift

I've turned the target object into a JSON schema, but you could probably generate that JSON schema pretty reliably using a codegen LLM.

benzguo··on Show HN: RAG on HN comments in 34 LOC
Ah, it's the "Preview" button in the val.town, or you can go to the hosted page: https://substrate-hackernewsrag.web.val.run
benzguo··on Show HN: RAG on HN comments in 34 LOC
Little demo that searches Hacker News comments for a topic (using https://hn.algolia.com/api), extracts sentiment and other metadata, then generates a research summary. Really proud of the API we've built at https://substrate.run – you don't have to think about graphs, but you implicitly create a DAG by relating tasks to each other. Because you submit the entire workflow to our inference service, you get automatic parallelization of dozens of LLM calls for free, zero roundtrips, and much faster execution of multi-step workflows (often running on the same machine).
benzguo··on Show HN: LLM draws SVG, StableDiffusion polishes it
i had this idea for an AI workflow:

1. describe an svg drawing in great detail

1.5. ask the llm to label its drawing because it's fun to watch

2. generate the svg

3. img-to-img with sdxl

it's a silly demo but kinda fun, source: https://www.val.town/v/substrate/scribble

you can coax Llama3 into generating better SVGs using multiple steps, actually – in a previous version the SVG had gradients, stacked layers, dot grid textures... even animations

but i had to do a roundabout thing to rasterize the SVG in the val sandbox so i removed all that

https://x.com/vprtwn/status/1811710646729277740

benzguo··on Show HN: Streaming Illustrated Primer in 150 LOC
The "Illustrated Primer" is a classic sci-fi vision for the future of education: dynamically generated educational content on any subject. It was surprisingly easy to put together a streaming demo on val.town (which is awesome).
benzguo··on Unit is a general purpose visual programming system
Second release since our public launch – we've been around in private beta for a bit and have had different versions of conditional support for a while!
benzguo··on LightRAG: The PyTorch Library for Large Language Model Applications
Neat, definitely aligned with this philosophy. If you like this, you'd appreciate Substrate - an even lighter weight, Pulumi-inspired API for modeling these workflows: https://x.com/vprtwn/status/1810363923750646032?s=46&t=HBHv3...
benzguo··on Unit is a general purpose visual programming system
Love unit, long time admirer. If you're into this, you might also be into Substrate - a non-visual but also graph-based programming model, for multi-step ML programs https://www.substrate.run/
benzguo··on Show HN: AI assisted image editing with audio instructions
Super cool! We're building an API that makes it easy to build chained multi-model workflows like this that run with zero latency between tasks - https://www.substrate.run/
benzguo··on An infinite canvas for code exploration
Funny coincidence – saw this very similar, perhaps more polished variant of the idea trending on twitter today: https://twitter.com/fronseneca/status/1752125155667824764?s=...
benzguo··on GitHub Copilot Workspace: Technical Preview
it's so over for Devin
benzguo··on Who is hiring interns (April 2024)
https://www.substrate.run/ - We're building the next-gen API for multi-model AI workloads. Cracked team, ex-Stripe, Substack, Pulumi. Sick office in east wburg, nyc
benzguo··on Ask HN: Who is hiring? (February 2024)
꩜Substrate | Founding Engineer | NYC – Hybrid/Onsite

We're building a never-before-seen way to declare, run, and scale an ensemble of ML models. You'll get to build a deep understanding of this fast-evolving space, design greenfield developer interfaces and infrastructure, and co-create the culture of what could become a generational company.

Because we're building an ambitious developer product, we're extremely selective with early hires. We're looking for engineers who are extraordinarily meticulous about the foundations, really sweat the details when designing interfaces, abstractions, and infrastructure, and are willing to wear the many hats we'll need to scale this business.

Our founders are friends that have known each other for the last decade since meeting at Venmo in 2013. Rob (CEO) was the founding engineer at Substack, Smalltalk, and Fin.com. He’s been hacking on GPU pipelines for years, and built AI features for startups before realizing he should be building reusable infra instead. Ben (Cofounder) spent the last 8+ years at Stripe, where he led mobile and was the founding engineer on Stripe Terminal. Our early team comes from Grailed, Stripe, Pulumi and the Samsung AI center.

We've raised a large seed round from notable VCs, the founders of Vercel, Venmo, Substack, and others.

Apply: Email ben@substrate.run with your resume

More info: https://publish.obsidian.md/-/Join+Substrate

benzguo··on Ask HN: Who is hiring? (January 2024)
꩜Substrate | Founding Engineer | NYC – Hybrid/Onsite

We're building a never-before-seen way to declare, run, and scale an ensemble of ML models. You'll get to build a deep understanding of this fast-evolving space, design greenfield developer interfaces and infrastructure, and co-create the culture of what could become a generational company.

Because we're building an ambitious developer product, we're extremely selective with early hires. We're looking for engineers who are extraordinarily meticulous about the foundations, really sweat the details when designing interfaces, abstractions, and infrastructure, and are willing to wear the many hats we'll need to scale this business.

Our founding team is a group of friends that have known each other for the last decade since meeting at Venmo in 2013. Rob (CEO) was the founding engineer at Substack, Smalltalk, and Fin.com. He’s been hacking on GPU pipelines for years, and built AI features for startups before realizing he should be building reusable infra instead. Ben (Cofounder) spent the last 8+ years at Stripe, where he led mobile and was the founding engineer on Stripe Terminal. Liam (founding engineer) spent the last 8 years at Grailed, recently acquired by GOAT.

We've raised a large seed round from notable VCs, the founders of Vercel, Venmo, Substack, and others.

Apply: Email ben@substrate.run with your resume

More info: https://publish.obsidian.md/-/Join+Substrate

benzguo··on Tarot for Techies
V2 cards! Thanks for the feedback everyone :) https://witchcraft.computer/major-arcana
benzguo··on Ask HN: Show me your half baked project
I’m working on a simple site builder, like Linktree but with styling and features more to my own taste. Not sure where it’s going long-term though: https://tray.club
benzguo··on Ask HN: Where do you regularly discover interesting links on internet?
I made this space to share cool links with friends: https://tray.club/ (inspired by communities like special.fish). Not the best place for link discovery yet, but I like sharing an evergreen bundle of links and notes with friends this way :)
benzguo··on Denmark to cull millions of minks over mutated coronavirus
this is genocide.
benzguo··on Show HN: Twitter threads to Markdown using the Twitter v2 API
Opening the garage door:

I built this over 2 days, Friday night and Saturday. Thanks to the new search endpoint in the v2 Twitter API, I was able to go from concept to launch really quickly. I definitely wouldn't have been able to build so quickly on pre-v2 APIs (and the user experience would've been wayy more janky).

Lots of positive feedback so far on Twitter, thanks to the #RoamCult: https://twitter.com/benzguo/status/1294832715049517058

I’m thinking about monetizing with a “top up credits” / pay-as-you-go pricing model. Would love advice from the HN crowd :)

benzguo··on Show HN: I built a Threadreaderapp alternative in 1 day using the v2 Twitter API
That'd be cool – I'll definitely consider it! Adding to the backlog :)
← PreviousPage 2 of 3Next →