Checked live against npm, GitHub, PyPI and IP India

Before You Build It,See What'sAlready There.

Aurexis checks your idea against live sources and shows you exactly what it found. You choose. Your choice is sealed.

Free, alwaysAbout a minuteNothing simulated
  1. 01Describeyour idea
  2. 02Check live4 sources
  3. 03Chooseproceed or halt
  4. 04Sealedon the record

Example check

npm

3 close matches

fetched 14:20 IST

Sealed · rec-8f2a41c9

Halt · condition set

Proceed if weekly downloads stay under 2k

Sealed · 14:20 ISThash 9f2a41c9…b7c14e
Recheck · day 30every source reruns
IP India · Unreachedour gap, never your evidence
aurexis.app/check/rec-8f2a41c9

Step 3 of 5   ·   What the live sources say

Here is the evidence. The call is yours.

0Supports
2Contradicts
2Absent
1Unreached
npmExact name taken. Maintained, 184k installs a week.Contradicts
PyPIExact name lookup returns 404.Absent
IP IndiaNo public feed. We could not ask.Unreached · our gap
Your callProceedHaltWait until

Every check reads live sources

npmPyPIGitHubUSPTOIP IndiaHacker News
How it works

Four steps. One record.

You bring the idea. Live sources bring the evidence. You make the call, and the record keeps it honest.

  1. Check

    We query live sources for your idea and show what each one supports, contradicts, or could not reach. Every finding carries its source and fetch time.

    about a minute
  2. Choose

    Proceed, Halt, or Wait until a date you pick. Your first choice is sealed and hashed the moment you confirm. It can never be edited.

    sealed and hashed
  3. Report

    The strongest case for your choice and the evidence against it. It never tells you that you chose well.

    case for and against
  4. Recheck

    At day 30 and day 90 every source reruns. You see what moved, and you can change course with the reason recorded.

    day 30 · day 90
What you get back

Evidence, not a verdict.

Every source answers in its own words, with the time we asked. Nothing is averaged into a score, and nothing is decided for you.

aurexis.app/record/rec-8f2a41c9Example recordchecked 14:20 IST

The idea

“An AI agent that triages and fixes GitHub issues for small teams.”

4 of 5 sources reached
2 Contradicts2 Absent1 Unreached
npmExact name taken. Maintained, 184k installs a week.Contradicts14:20
GitHub41 repositories match. 3 active in the last 90 days.Contradicts14:20
PyPIExact name lookup returns 404.Absent · a finding14:21
USPTONo conflicting mark found in class 9.Absent · a finding14:20
IP IndiaNo public feed. We could not ask, so we cannot say.Unreached · our gap3 retries
Your choice

Three answers. All of them respected.

Whichever you pick, the report sets out the strongest case for it and the evidence against it. It never tells you that you chose well.

Proceed

Build the smallest slice, with the conditions that would make you stop.

Your report includes

  • The case for going, with sources
  • The evidence against, stated plainly
  • Halt conditions in your own words
  • Rechecks at day 30 and day 90
then the Blueprint, if you want it

Halt

Stop here, and keep what would change your mind.

Your report includes

  • The case for stopping, with sources
  • The evidence against halting
  • What would turn it into a Proceed
  • We watch those signals for you
halting is always free

Wait until

Not yet. Pick a date or an event to revisit on.

Your report includes

  • What is unresolved today
  • The evidence against waiting
  • Every source rechecked on your date
  • No date means a wait becomes a halt
a date or a trigger is required
Your line

You draw the line.
We measure it.

When you proceed, you write what would make you stop, in your own words. On each date we read the live source and mark it Held, Broke or Unknown. No rulebook, in any domain.

Held

Your line was not crossed.

Broke

Your line was crossed. You decide what happens next.

Unknown

Source unreachable after three retries. Never counted as held.

Your halt conditionsExample

“If fewer than 20 teams sign up in the first 30 days, I stop.”

Held
stop zoneyour line · 20 teams
by day 30 · 43 teams0 to 60

“If day 30 retention is below 40%, I stop.”

Broke
stop zoneyour line · 40%
by day 30 · 31% retained0% to 100%

“If a project in this space passes 5,000 GitHub stars before March, I stop.”

Watching
stop zoneyour line · 5,000 stars
by 31 Mar · 2,104 now0 to 6,000

Written by you. We never move the line.

Same methodAny domain
Kirana store app

“If fewer than 5 shops are paying by 31 March, I stop.”

Developer tool

“If weekly npm downloads fall below 2,000 for four weeks, I stop.”

Grant led biotech

“If the BIRAC call closes without a shortlist, I stop.”

Method

Not every empty result means the same thing.

Most tools merge two very different outcomes into one blank. We keep them apart on every check.

We ask a live source

Every check, every source, every time. Nothing cached, nothing guessed.

Where other tools show a blankAbsent and Unreached look the same elsewhere. Here they never merge.

Did it answer?Yes
No
Unreached

Could not ask

The source did not answer. Our gap, never your evidence. We retry three times, then say so plainly.

IP India no public feed
Did it find anything?Yes
No
Absent

Asked, found nothing

The source answered with nothing. That is a finding, and it counts as evidence.

PyPI exact name, 404
Supports

Backs part of it

A source supports some part of the idea.

Contradicts

Cuts against it

A source shows the idea, or close to it, already exists.

Every state links to the source that produced it.

Rechecks

Your choice is sealed.
The evidence keeps moving.

We rerun every source on a schedule, so you can see what changed since you decided.

Example recordOne idea, ninety days
Day 30 recheck is in1 source moved since you decided.
See what moved
  1. Day 0Sealed

    Your call: Proceed

    Evidence frozen. First choice hashed and timestamped.

    5 sourcesFrozen
    record9f2a41c9…b7c14e
  2. Day 301 moved

    Recheck

    Every source reruns. Conditions due by now are read.

    PyPIAbsentContradicts
    halt conditionHeld
  3. Day 90Nothing moved

    Recheck

    Everything reruns again. You see what moved, and what did not.

    5 sourcesUnchanged
    halt conditionUnknown

Three retries, then Unknown.

If a source will not answer, we say so. We never fill the gap with a guess.

Change course any time.

The change sits next to your first choice, with the reason: new evidence, or none.

The ledger

The Ledger counts decisions, not people.

Totals only. No ideas, no names, no single records. Every number comes from a sealed first choice and the halt conditions founders wrote themselves.

Counts before percentages.A rate appears only once there are enough decisions behind it.
Unknown is never folded into Held.An unreachable source is our gap, counted on its own.
Your idea is never shown.Your record stays yours to share or not.
Open the Ledger
aurexis.app/ledgerCounting from 11 Nov 2026
First choices
0sealed
Proceed 0Halt 0Wait 0
Changed later
0changed
After new evidence 0With none 0
Halt conditions due
0reached date
Held 0Broke 0Unknown 0

No decisions yet. The first sealed choice starts the count.

How a total reads
3 of 4 held
× 75%

Too few decisions for a rate to mean anything.

What we refuse to do

Six things
we will never do.

Most products promise what they will do. These are the promises that make a record worth keeping.

01Never

Issue a verdict

Instead

The sources speak. You decide.

02Never

Turn evidence into a score

Instead

Every finding stays in its own words.

03Never

Hide a source we could not reach

Instead

Unreached is shown, every time.

04Never

Edit your first choice

Instead

Sealed and hashed the moment you confirm.

05Never

Move your line

Instead

Your halt conditions are yours alone.

06Never

Put your idea on the Ledger

Instead

The Ledger shows totals only.

Your idea, checked in about a minute

See what is already there.

↵
Free, alwaysAbout a minuteNothing simulated

Check it without us.

Every record is plain JSON. Canonicalise it, take the SHA-256, compare. No account, no network. If our servers disappear tomorrow, every record stays checkable.

  1. 1Canonicalise
  2. 2SHA-256
  3. 3Compare
Verify a record
record.json · example

$ jq -S -c . record.json | sha256sum

9f2a41c93e8b0c2d77f91b5e6a04c8d3f2e1b9a7c6d5ef4f30812a6b4cb7c14e

# same hash as the one printed on your record

Match. This record has not changed.