Product updates · last shipped 31 Aug 2026
What changed, and why it matters.
New screens, new rules, changes to how contracts are scored, and the work underneath that you would otherwise only notice when it breaks.
August 2026
Watchlist
Save any company from a scan and it keeps what the screen said about it that day — price, fair value, upside, score, dividend safety and capitulation. Add the same ticker again and the snapshot updates rather than duplicating.
Why it matters: A scan is a moment in time. The useful question a week later is not what the name is worth now, it is whether what you believed then still holds — which needs the old numbers sitting next to the new ones.
Dividend safety score
A 0-100 reading on every dividend payer, from free cash flow coverage, the earnings payout ratio, leverage, and how far the yield sits above the market's. The arithmetic is shown in the company drawer, and a score under 40 raises a risk flag.
Why it matters: Yield tells you what a stock pays today; it says nothing about whether the business covers it. Coverage here is measured against cash rather than earnings, because the dividend is paid in cash. Companies that pay nothing score nothing at all rather than zero — no dividend and an unsafe dividend are different answers.
Capitulation score
RSI, drawdown from the 52-week high, and recent volume against the year's own average, combined into a 0-100 reading on every match. The Capitulating flag needs all three at once; any one alone is just a chart.
Why it matters: Cheap and still falling is not the same as cheap. A company can pass every balance-sheet rule while the selling is accelerating, and knowing which of those you are looking at changes when you act, not whether.
Run screens from an AI assistant
A Model Context Protocol endpoint at /api/mcp exposes the screener as two tools: run a screen, and list every metric and threshold. Point any MCP-capable assistant at it.
Why it matters: It authenticates as you and spends a credit exactly like the browser does, so an assistant gets your access and your quota and nothing more. Asking a question in prose beats translating it into eleven threshold fields first.
Every screen and metric documented in public
Each of the seven screens now has its own page listing every rule it applies and why. There is a glossary defining all 28 metrics the engine tests, and a set of articles on net-nets, EV/EBIT against EV/EBITDA, buying LEAPs on fundamentals, and where screener data is wrong.
Why it matters: The rules and the reasoning were only visible after signing in and running a scan. Deciding whether a screener thinks the way you do should not require paying for it first.
The screener and the engine ship as one container
The Bun server and the Python screening engine now build into a single image and deploy together, rather than the UI calling a separately hosted engine.
Why it matters: One deploy instead of two, and no network hop between the page you are looking at and the process running the scan.
The engine moved behind an HTTP endpoint
Scans are driven over a plain HTTP API that streams progress and results back as the Python process produces them, instead of the UI shelling out directly.
Why it matters: Progress arrives while the scan is still running, a scan can be stopped mid-flight, and the front end is no longer tied to running on the same box as the engine.
First release
Seven screens — NCAV, negative enterprise value, the Acquirer's Multiple, the Simple Way, the combined Core Deep Value gate and two LEAP strategies — over a universe you bound by market cap, volume, sector and country. Every rule reported with the value it was tested against, near misses listed with what they failed and by how much, option contracts scored with implied volatility and delta computed from the chain, and analyst notes written from the scan output.
Why it matters: Everything the screener claims is a rule you can read and a threshold you can change, rather than a score you have to take on faith.