elawler24

This reminds me of Cloudflare's focus on the innovator's solution and how they justified competing against giants like AWS and Google. They created a new market focused on startup adoption, based on ease of adoption and more reasonable costs (free egress fees on R2 for example, to reduce data lock-in concerns).

Detailed article from Stratechery here that's worth a read - https://stratechery.com/2021/cloudflares-disruption/

Before, I ate low sugar / carb and high on cheese, meat, whole milk, yogurt, and veggies (close to keto). Now I eat a lot of rice, beans, while grains, and veggies. I’m trying to figure out how to get enough protein though, that’s the trade off.

I do have a fear that staying vegan will have other negative effects, unrelated to cholesterol - esp brain and bone density related. For me that means cutting out cheese, milk, and butter as much as possible. But having fish and lean meats 1-2 times per week.

High cholesterol and heart attacks are common in my family. This year, after my dad had intensive open heart surgery, my doctor recommended trying a strict plant-based diet for 90 days with a blood test before and after. She had been studying medical journals on the topic primarily from Canada (she said it’s easier to find medical research not funded by corporations there).

Before the doing the plant-based diet, I had such high cholesterol that I would have needed to start taking statins before age 35. After the 90 day diet experiment, my cholesterol dropped by 130 mg/dL. I no longer need to be put on medication, and am within a healthy range.

The best way to get started is to use a trusted design system on the frontend, and follow their components, layouts, and color scheme exactly. Once you know the rules and have something basic implemented, you can break the rules from there. I think most developers go wrong when they try to re-invent the wheel before inherently understanding the rules (by copying experts). Tailwind, Google Material Design, and Apple UI interface are good starting points that are well documented.

It’s a grind to run this kind of business. You’re sometimes dealing with life and death situations (you’re liable for someone’s electricity or heating 24/7). Talent is really hard to come by at a reasonable rate, especially reliable master electricians or plumbers and the trades people that work under them. And then it’s an extremely cost and labor intensive business involving large machinery, equipment, and vehicles - inside people’s homes. (Watched my dad scale one of these businesses for 30+ years, his work was stressful and unending)

My dad bought a failing HVAC business 30+ years ago, then made it profitable over the years and sold it back to his employees last year. He had the option to take a few highly lucrative PE deals, but it was clear they would squeeze the life out of the employees and customers he had worked hard to support over many years. I can’t imagine how low quality this kind of trade work will become if PE owns them all. It will be similar to vet, dentist, and dermatology clinics which now feel like factories that don’t care about the humans on the other end - often employing fear tactics and sales quotas to incentivize upsells.

It’s not simply more hours in, more money out. It’s the desire to work hard that matters.

This is the difference between junior investment bankers and engineers. Both have to work hard at the beginning of their careers to justify being paid for their work for all future years. Investment bankers are paid to be available to make decks, crunch numbers, etc at all hours of the day. I’m curious how many of those hours are spent doing something of value vs just being available to the partner. In comparison, a software engineers time is not valued for simply being available, but rather in creating something useful. Both are motivated to work many hours, but one is intrinsic and the other is extrinsic.

Thank you for trying to take on this problem, it’s really bad though. I did a PH launch this summer resulting in a few hundred unique website views. Comments besides people I already knew were either bots or promotional. Lots of bounty hunters.

Compare that to our HN post requiring no additional lift, which resulted in ~10k unique visitors, no bots, and great critical feedback we couldn’t get anywhere else.

I do want an authentic PH! But it feels like a vanity landing page in its current state.

Got it. We can store logs to your purpose-built analytics DB of choice.

PostgreSQL (Neon) is our free self-serve offering because it’s easy to spin up quickly.

It's a standalone DB, just for LLM logging. Since it's your DB - you can configure data retention, and migrate data to an analytics DB / warehouse if cost or latency becomes a concern. And, we're happy to support whatever DB you require (ClickHouse, Big Query, Snowflake, etc) in a managed deployment.

We can support any database you need, PostgreSQL is the easiest way to get started.

I’m a product designer, so we tend to approach everything from first principles. Our aim is to keep as much complexity in code as possible, and only surface UI when it solves a problem for our users. We like using tools like Vercel and Supabase - so a lot of UI inspiration comes from the way they surface data views. The AI phase of the internet will likely be less UI focused, which allows for more integrated and simple design systems.

We already support OpenAI and Anthropic endpoints, and can add models/endpoints quickly based on your requirements. We plan to expand to Llama and other self-hosted models soon. Do you have a specific model you want supported?

We could support any API. We’re focused on building data pipelines and tooling for LLM use cases.

Right now, it’s a managed service that we set up for you (we’re still a small team). Email me if you’re interested and I can share details - [email protected].

Velvet takes <5 mins to get set up in any language, which is why we started as a proxy. We offer managed / custom deployments for enterprise customers, so we can support your client requirements.

We warehouse logs directly to your DB, so you can do whatever you want with the data. Build company ops on top of the DB, run your own evals, join with other tables, hash data, etc.

We’re focusing on backend eng workflows so it’s simple to run continuous monitoring, evals, and fine-tuning with any model. Our interface will focus on surfacing data and analytics to PMs and researchers.

For pgvector/pinecone RAG traces - you can start by including meta tags in the header. Those values will be queryable in the JSON object.

Curious to learn more though - feel free to email me at [email protected].

Might be a Cloudflare flag. Can you email me your IP address and we'll look into it? [email protected].

Thanks! You can layer on as much detail as you need by including meta tags in the header, which is useful for tracing RAG and agent pipelines. But would love to understand your particular RAG setup and whether that gives you enough granularity. Feel free to email me too - [email protected]

Depends on the scenario. In a threaded query, or multiple queries from the same user - you’d want different outputs. If 20 different users are looking for the same result - a cache would return the right answer immediately for no marginal cost.

Thanks for the detail! This is a use case we plan to support, and it will be configurable (for when you don’t want it). Some of our customers run into this when different users ask a similar query - “NY-based consumer founders” vs “consumer founders in NY”.

There were plenty of UI-based low code platforms. But they required that we adopt new abstractions, use their UI, and log into 5 different tools (logging, observability, analytics, evals, fine-tuning) just to run basic software infra. We didn’t feel these would be long-term solutions, and just wanted the data in our own DB.

The self-serve version is hosted (it’s easy to try locally), but we offer managed deployments where you bring your own DB. In this case your data is 100% yours, in your PostgreSQL. That’s how Find AI uses Velvet.

Hey HN! We’re Emma and Chris, founders of Velvet (https://www.usevelvet.com).

Velvet proxies OpenAI calls and stores the requests and responses in your PostgreSQL database. That way, you can analyze logs with SQL (instead of a clunky UI). You can also set headers to add caching and metadata (for analysis).

Backstory: We started by building some more general AI data tools (like a text-to-SQL editor). We were frustrated by the lack of basic LLM infrastructure, so ended up pivoting to focus on the tooling we wanted. So many existing apps, like Helicone, were hard to use as power users. We just wanted a database.

Scale: We’ve already warehoused 50m requests for customers, and have optimized the platform for scale and latency. We’ve built the proxy on Cloudflare Workers, and latency is nominal. We’ve built some “yak shaving” features that were really complex such as decomposing OpenAI Batch API requests so you can track each log individually. One of our early customers (https://usefind.ai/) makes millions of OpenAI requests per day, up to 1500 requests per second.

Vision: We’re trying to build development tools that have as little UI as possible, that can be controlled entirely with headers and code. We also want to blend cloud and on-prem for the best of both worlds — allowing for both automatic updates and complete data ownership.

Here are some things you can do with Velvet logs:

- Observe requests, responses, and latency

- Analyze costs by metadata, such as user ID

- Track batch progress and speed

- Evaluate model changes

- Export datasets for fine-tuning of gpt-4o-mini

(this video shows how to do each of those: https://www.youtube.com/watch?v=KaFkRi5ESi8)

--

To see how it works, try chatting with our demo app that you can use without logging in: https://www.usevelvet.com/sandbox

Setting up your own proxy is 2 lines of code and takes ~5 mins.

Try it out and let us know what you think!

This sounds like the engineer / designer dream. Make the best product possible, and the users will follow. Vibes work in a world where money is free and timelines are endless (low interest rates, funding the business with contract work, lots of angel investors, side projects, the innovation team).

It’s also a function of what phase the product is at. Do you have users yet? Do they use the product yet? Do they stay retained? What features matter to them? All of those questions are very quickly answered with data. They help align a group of collaborators on what matters.

Eventually it becomes a requirement for closing deals with companies who have compliance requirements with their own customers. But it’s expensive and time consuming, so only worth it if it’s actually blocking large contracts. You can use something like Vanta when you’re ready.

I can move faster in Figma than most people can type the NL query to automate code. Which helps us validate high-level ideas before it “works”. It forces us to take a beat and make sure we’re introducing something of value to the internet and our users. But maybe we never need to worry about maintaining components or grids or pixels again - the robots can do that.

But all of the example stories are far deeper than “wrappers”. They do have a moat, and use OpenAI (or x model) as a commodity like AWS or Google Cloud. They’re mixing models with their proprietary data and leveraging techniques that are hard to copy (either due to cost, time, or traditional network effects).

If you’re bootstrapping a company to make personal money - buy ads and be a wrapper. If you want VCs, companies, and consumers to buy in long-term - you still need depth. Even in the world of AI, that comes from owning data + network effects, not just an AI interface.

Fundraising for the first time might be hard - persistence and tenacity will get you through it. Put together a pitch deck, demo, maybe a 1-pager, and a spreadsheet of every founder and investor you know. Start lining up meetings, and your pitch will get better every time. You might do hundreds of them. Your first check will likely come from a personal connection, then it will snowball from there.

Take checks as SAFEs - YC has a standard one you can download that everyone trusts. I would watch all the YC startup school videos on fundraising if you haven’t already, lots of great advice in there.

It only takes 1 yes for a round to come together.