Here's a 2.5 min demo if you want to skip ahead: https://www.youtube.com/watch?v=g_E6yhpN-zU

I've been using coding agents for a few years and have been frustrated by how quickly they jump to solutions. Even with plan mode and ask-user-question features, it's still a review dynamic. You describe a problem, they hand you a complete solution, you react. The solution looks plausible but may hide bugs or design flaws you can't catch, because you didn't think through the design yourself. Over time, your intuition for your own software degrades because you've stopped making meaningful decisions.

rubberduck flips this. Each session has three phases: design, implementation plan, and execution.

In the design phase, you describe what you want to build and rubberduck chats with you about how to solve it. I was trying to mimic the design conversations I used to have with colleagues at Microsoft and Stripe: sitting down, talking through options, poking holes in them, weighing trade-offs. rubberduck explores the design with you and investigates the codebase on your behalf, but you make every relevant decision. The output is a design document.

Once you're happy with that, the implementation plan phase takes over. The agent does a deep dive into the codebase to figure out how the design maps to actual code. This is similar to plan mode in Claude Code. It's mostly autonomous, since the meaningful decisions are already made.

Finally, execution. The agent runs in an isolated container using a dev environment image you provide. Each task is validated against your quality gates (tests, linting, formatting) before being committed. The output is a PR.

Pricing is $24/month with $5 of included credits (roughly one session, depending on complexity). Credits cover tokens (no markup) and execution compute (you specify the vCPUs/memory you need).

I've been bootstrapping for a few months and it's been fun to build. Would love for you to try it. Feedback very welcome.

Comments URL: https://news.ycombinator.com/item?id=48065297

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=21114744

Points: 3

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=20215292

Points: 23

# Comments: 18

5% is $0.5 which right off the bat assumes payment processing is separate. That's fine if people prefer that but even then $0.5 is barely enough to cover infrastructure. How do we pay for someone to maintain that infrastructure? This is the same issue OSS is facing but on the other side. No one wins when the tools we rely on don't have a sustainable business model. I don't think 5% is sustainable as we've seen with Patreon's recent increase.

>Personally i am looking for a way to give to project i use the most, (firefox, vim , zsh etc...) and also project that i use and particularly need funding.

Yea this mostly geared towards libraries and frameworks not apps.

>Any plans for the entreprise market ?

Enterprise is an important market but I was hoping to get some early adopters to prove the idea first and then use that proof to get businesses to join. Not opposed to doing both in parallel though.

30% includes all payment processing costs and other costs for both users and projects, so there is a flat 70% remaining for projects that doesn't get deducted further. We do this to keep things simple and make it clear how much projects can get.

Payment processing alone costs $0.3 + 3% with Stripe, so at $10/month that's a fixed $0.6 out of the $3 we take. The other $2.4 needs to accommodate the fees to send money to each of the projects + infrastructure costs. Not sure how to account for that but assuming at scale around $0.4/user/month, we're looking at $2/user/month in profit. Assuming sustainability is reached at $200,000, that is approximately 8300 people giving to open source through this. That is high but seems approachable.

Does that seem reasonable? My objective with this is to build something that increases the pool of people who contribute but also ensure the underlying project can stay afloat indefinitely with the right incentives.

The problem with this mechanism is that now each project is out on their own to advertise their needs and get donations. I'm actually experimenting in this space with SeedAndDew (https://www.seedanddew.com). My hypothesis is that if you can give people the ability to set a specific amount to open source that automatically goes to the right places, more people will contribute. Just went full-time to see this through so early days but I'm optimistic.

I think people are just getting hung up on the number without thinking deeper on what it costs to make things sustainable.

5% of the 500 million Patreon claims to be able to process is 25 million. According to their team picture on their website (https://www.patreon.com/about), they have 74 employees. That amounts to $330000/employee. Halving that to account for benefits (chair, table, wifi, insurance, etc) leads to $165000/employee for salary which seems reasonable average for a senior employee in the Bay Area. Notice this doesn't account for infrastructure, taxes, outside counsel/contractors, etc. It's crazy that after working for 6 years and creating half a billion of value, a company can barely afford to keep the lights on, let alone make a profit. I agree 30% just makes sense for a platform - you need enough to hire properly, pay all expenses and make enough of a profit for this to make sense. The alternative is half a billion of value disappears to save on a few million which I think is incredibly myopic.

Comments URL: https://news.ycombinator.com/item?id=18295600

Points: 6

# Comments: 1

The solutions I've considered are DataDog and AppDynamics but they seem to focus on the server health instead of the application logs which is what I want.

I would prefer not to build my own solution here. Is this a solved problem?

Comments URL: https://news.ycombinator.com/item?id=17207796

Points: 1

# Comments: 4