juecd

firecracker is an obvious place to start because it offers best-in-class support fast cold starts and hard isolation. we can't speak to what exactly the other browser infra providers are doing, but based on public benchmarks we're up to 5.8x faster

As part of launching our MCP server, we wrote a short blog post about some of our technical decisions and limitations of the current landscape. The space still feels very early, so would love others' thoughts on the MCP space and what else is needed / how you think it'll evolve over time.

Thank you! Just saw your GH, will check out your WebRTC work more deeply :)

Oops, I forgot to add the link: https://github.com/onkernel/kernel-images

The big motivation we had for trying to do this on a unikernel is actually the session pause/resume. The fast cold restart is a nice addition (or put the other way, a super slow restart would prohibit session pause/resume from being useful)

Yes

Are you referring to Unikraft Cloud's pricing? I'm not really sure, since I'm not affiliated. Maybe they'll chime in

Thank you!

Thank you! Yeah, the current implementation basically performs the same as a Docker container (i.e. not much). The interface is the same, so you can use BU/Playwright/Puppeteer's header configs to change as needed.

We did notice the unikernel cloud instances don't run into bot detection as often as our hosted docker instances, but I think that's mostly because Cloudflare haven't flagged Unikraft Cloud's IPs yet, hah.

From what we've seen, micro VMs could probably do something very fast too (150ms?) but we thought 20ms was pretty crazy.

Thank you!

(OP) Happy to answer any questions! Some things we're still exploring:

- Mounting persistent storage for file i/o

- Replacing noVNC with a faster alternative

Would love feedback, especially if you’ve worked on fast-cold-start systems or unikernel-based infra.

Cool concept - what sorts of tasks are you hoping to do in the future?

Your demo is awesome, nice job. Do you guarantee a certain response time?

Also, I'm curious how people use your API if it takes 30 seconds+ to get a response -- ie can I use this for time-sensitive annotations, and also, wouldn't my code hang waiting for your response or do you offer webhooks?

Super good point that we've found with our customers who are migrating from outsourced bookkeepers.

We're using our Slackbot as a friendly way for companies to get their financial data, the idea being that if the numbers are accessible/convenient and ones they want to know (revenue, burn rate, bank balances). Then, we can use the same Slackbot to provide other necessary financial details.

We're still working on this, but we think Gusto does a great job of communicating necessary (and not so exciting) info in a way that engages us.

Xero and Quickbooks just use basic regex to match transactions, which doesn't make much sense if the transaction category isn't static, eg Amazon. We also learn categories on an aggregate level, so if your bank statement's never seen a transaction from, say Tommy's Tacos, but someone else has, we'll be able to categorize it correctly immediately.

For categorization, we're using ML (which learns over time and isn't rule based like Quickbooks' suggestions, which uses regex).

For the other pieces of bookkeeping, we're using integrations to automatically match invoices/payments (especially disjointed ones, eg invoice through QBO but pay separate via Stripe on client's website). We're also doing stuff like adjusting journal entries programmatically, like payroll accruals. For now, QB is providing the financial reports, we're doing the bookkeeping so their automated reports are accurate!

We're basically doing the services that a bookkeeper would provide on top of a client's Quickbooks software. QB is a tool and we aim to be a service.

Right now, Quickbooks Online only. We can help switch to QBO if you're open to it!

Glad to know auto revenue recognition is useful to you! Not yet, but Braintree is high on our list of integrations to build

Thanks so much!

We're using machine learning (supervised learning) to categorize transactions, generally expenses. We're using our users' past months' books as our training sets of correctly labelled data, so that we can use that data to train is really exciting for us!

Hey guys! One of the co-founders of Sway Finance here. We did a ShowHN a couple months ago when we had just built our Slackbot (https://news.ycombinator.com/item?id=11162283) and are excited to now be in YC and building more full-fledged financial products for businesses. Happy to answer any questions here!

Super excited to try this — as an early start-up (and the founder in charge of UI), I KNOW our site has UI issues but I don't have the time to do full inspections.

What kind of "changes need attention" will Spotbot tell me about?

We also allow email notifications but figure most teams prefer to integrate through Slack.

Physical location and about the team is a great point. Appreciate it!

Great point about the privacy policy. It's important to us to make it clear that we're not misusing the information. Thanks for the feedback!

Got it - appreciate the feedback!

Makes sense. We're starting with accessible (and push-based) information to complement founders' existing systems and then planning to build upon it in the future. Thanks for the feedback!

We're only in the US for now (most financial institutions supported). Hoping to expand access in the future!

We made a Slackbot for startups to tell them some simple financial metrics (how much is in their bank account, and what their latest transactions in those accounts are). We're connecting to company accounts through Plaid (plaid.com) and sending them a financial 'digest' through Slack/email each morning.

As consumers we've been using a couple apps like Digit (digit.co) to track our personal finances but couldn't find a push-based version for companies. Logging into our company bank accounts to keep track of transactions has been pretty painful (online business banking UI is really hard to navigate) so hopefully this can make the process easier.

We want to eventually incorporate burn-rate to help founders stay on top of their company's money. Would love any feedback -- thanks!