Yes or no
Noulwaiting for the state…
you've heard the hype
Now see how jevify actually works.
Eighteen real messages, scored by the active decision provider as this page opens. Drag the thresholds to see what falls in or out.
jevify sandbox
Play with every threshold, curve and zone. Nothing to install.
curves · zones · heat map · simplex · 3D · migration
three typed decisions
One state. Three bounded answers your code can use directly.
state · the input jev reads
typing…“
Pick one
Choicewaiting for the state…
Rate a level
Scorewaiting for the state…
visual classification
Move one limit. Watch every item move.
60 simulated itemsOpen the Sandbox
share of items
rejectscore 0–0.40
20%12 items · left out
reviewscore 0.40–0.75
40%24 items · human check
acceptscore 0.75–1
40%24 items · code acts
accept threshold· auto-sweeping, drag to take over0.75
more automatedmore selective
Item distribution
Horizontal position = score. Color = current zone.
reject · 0–0.40review · 0.40–0.75accept · 0.75–1
0.00decision score1.00
decision field
18 messages, one decision call each, scored on 4 categories plus severity and escalation.
Inside
4
Review
1
Outside
13
stored run
Billing · charges, invoices, refunds and pricing
How the scoring works
- 1 · ask — each message goes out once with three typed questions: which team owns it, how severe it is, and whether a human is required.
- 2 · score — the answer comes back as probabilities per team, a 1–4 severity and one escalation probability. No prose.
- 3 · decide — your three limits above turn those numbers into act, review or reject. Move a slider and every dot re-zones instantly.
Nothing is waiting on a human here: the limits, not the model, decide what gets automated.
JEV decision run · model typesafe/jev-1.13-20260917 on Lovable AI Gateway · whole batch in 1140 ms · 8188 input tokens · billed 0.000344
Every value comes from the named provider. OpenRouter presets simulate this contract; TypeSafe routes execute JEV. Nothing here measures your own traffic.
Accept
0.80
Review
0.35
Human gate
0.70
0.000.250.500.751.000.000.250.500.751.00category probability →↑ human escalation probability
Inside · cat 100% · esc 49% · sev 1.7
dot size = severitylive provider results · 18 items
cell by cell
18 rows · live
| message | Bill | Tech | Acco | Feed | sev | zone |
|---|---|---|---|---|---|---|
| I was charged twice for the November invoice and the second charge is still pending on my card. | 1.00 | 0.00 | 0.00 | 0.00 | 1.7 | |
| The Stripe webhook has been returning 500 since Friday and no order is being confirmed. | 0.00 | 1.00 | 0.00 | 0.00 | 2.8 | |
| I cannot log in anymore. The reset email never arrives, I checked spam twice. | 0.00 | 0.04 | 0.96 | 0.00 | 2.0 | |
| Honestly the new editor is much nicer than the old one. Would love a dark mode for the sidebar too. | 0.00 | 0.00 | 0.00 | 1.00 | 0.0 | |
| My card was declined and now the workspace is read-only, my whole team is blocked mid-launch. | 0.94 | 0.03 | 0.03 | 0.00 | 2.2 | |
| Quick question: does the Pro plan include the audit log, or is that enterprise only? | 0.90 | 0.00 | 0.10 | 0.00 | 0.0 | |
| Someone from another company shows up in my members list and I have no idea how they got there. | 0.00 | 0.12 | 0.88 | 0.00 | 0.9 | |
| The export button does nothing on Safari. Works on Chrome. No error, just nothing happens. | 0.00 | 1.00 | 0.00 | 0.00 | 1.8 | |
| Please cancel my subscription and refund the last two months, I stopped using it in September. | 1.00 | 0.00 | 0.00 | 0.00 | 2.5 | |
| Not a complaint, just noting the docs page for webhooks has a broken link in the retry section. | 0.00 | 0.75 | 0.00 | 0.25 | 0.0 | |
| We lost three hours of submitted forms during your maintenance window and the data is gone. | 0.00 | 1.00 | 0.00 | 0.00 | 3.0 | |
| Can you move my seat to the new admin email? Same person, new domain after the acquisition. | 0.00 | 0.00 | 1.00 | 0.00 | 0.7 | |
| The invoice PDF downloads blank since the last release, and the amount on screen is different from the email. | 0.21 | 0.79 | 0.00 | 0.00 | 2.6 | |
| Our SSO stopped accepting new members right after we upgraded the plan. Not sure which change caused it. | 0.02 | 0.78 | 0.20 | 0.00 | 2.0 | |
| I was told usage-based pricing would be clearer in the dashboard. It still shows one number with no breakdown. | 0.81 | 0.03 | 0.00 | 0.16 | 1.1 | |
| Two people on the team see different totals for the same report. One of them is billed as a viewer. | 0.32 | 0.27 | 0.41 | 0.00 | 1.8 | |
| The trial ended while a sync was running. Now the integration is half connected and I cannot retry it. | 0.00 | 1.00 | 0.00 | 0.00 | 2.0 | |
| Love the API, but rate limits are not documented anywhere and we only found them by getting blocked. | 0.00 | 0.40 | 0.00 | 0.60 | 1.4 |
Inside
4
Review
1
Outside
13
01 · MCP
Run the MCP
Connect an MCP client and run jevify against a repository or supplied files. The audit is read-only and returns structured findings.
OAuth/mcpread-only
Run MCP02 · Skill
Download the app transformation skill
The jevify skill audits your application and migrates only the AI call-sites where Jev clearly delivers better performance or lower cost. Everything else stays as it is: generation stays generation, rules stay in code. Lovable is one of the runtimes it supports.
transformationselectivereport
scope
Use LLMs for language, code for rules, and JEV-style calls for bounded judgment.