TLDRocket
Sign in

Launch HN: Prized (YC S26) – Let non-engineer staff build secure internal tools

prized.dev marinoseliades

A YC startup called Prized lets non-coders describe an internal tool and get a working app hooked to company data. It's basically a way to turn stray spreadsheets and notebooks into real software without bugging engineering.

Every company has that one spreadsheet or Jupyter notebook doing load-bearing work nobody officially approved. Prized, a new YC S26 startup from founders Marinos and Hudson, wants to turn those orphaned scripts into proper internal tools without routing them through an engineering backlog that never gets shorter.

The pitch is simple: an employee describes what they need in plain language, and Prized generates a full-stack app wired to the company's actual data, sitting behind existing single sign-on. One customer's data scientist reportedly pasted in a personal fraud-detection notebook, hardcoded thresholds included, and after a handful of prompts it became a shared risk console with those thresholds exposed as adjustable UI controls. The founders say most of that company now uses it.

The harder problem, and the one Prized seems to actually care about, is keeping this from turning into a security nightmare. The build sandbox never touches real credentials. It works with scoped session tokens that are just placeholders; actual secrets get swapped in later at an egress proxy, and any outbound call to a connector passes through an LLM judge meant to catch dangerous requests before they happen. Network access defaults to deny, so nothing leaves the sandbox except through that proxy. Each generated tool also gets its own Postgres schema and role, with queries running through an authenticated SQL gateway, and admins can scope who sees what, with every data access logged.

What sets Prized apart from something like Lovable or Retool is the social layer bolted on top. Tools aren't private projects; they're shared objects other employees in the workspace can see, fork, and remix. The example the founders give is a marketing analytics tool that a data scientist later forked to add confidence intervals, without starting from scratch. The bet is that a workspace full of these forkable tools becomes something closer to an internal software library than a pile of one-off scripts.

Pricing starts free for two tool builds a month, with a $100/month Teams tier and a custom Enterprise plan for things like on-prem deployment. The founders are candid that they're still tuning the line between letting people build freely and making sure that freedom doesn't turn into unaudited access to sensitive systems.

My take

Letting non-engineers spin up tools tied to real company data is exactly the kind of thing that sounds great until an LLM judge misses one dangerous call and someone exports a customer table they shouldn't see. The security architecture here is genuinely more thoughtful than most vibe-coding startups bother with, but the whole category is basically betting that AI review layers can substitute for the code review and access controls engineering teams built for good reasons. Worth watching whether enterprises buy that trade, or whether this ends up as shadow IT with a nicer UI.

Read more about this at: prized.dev

Related stories

The daily briefing

Every AI story that matters, in your inbox by 8am.

TLDRocket reads all relevant sources, removes duplicate coverage, and summarises the day in two minutes. Follow companies and topics for alerts, or get the briefing in Slack. Free, no spam, unsubscribe anytime.