A plain-English guide

What it does

• Format — takes a wall of text squashed onto one line and lays it out neatly so a human can read it.
• Validate — tells you if the file is broken (a missing bracket, a stray comma) and where.
• Compare — shows two versions side by side and highlights exactly what changed.
• Explore JSON — opens a big data file as a foldable tree, and lets you search inside it for one value.
• Simulate — for HTML/CSS/JS, shows a live preview of what the code produces.

When you'd use it

• An engineer sends you an API response or config as one long unreadable line and you need to make sense of it.
• You want to see what changed between an old and new version of a file.
• Something isn't working and you want to check the file isn't simply broken.

What it does

• Combine several PDFs into one, or split pages out of a PDF.
• Compress a heavy PDF so it's small enough to email or upload to a portal.
• Add page numbers or a watermark.
• Black-out (redact) sensitive text so it's genuinely removed, not just hidden.
• Read a scan (OCR) — turn a scanned image into selectable, copyable text.
• Resize / convert / shrink images — e.g. get a photo down to an exact size a portal will accept.

When you'd use it

• A portal rejects your file for being too big — compress it.
• You need to merge several documents into one PDF to attach.
• You must hide an account number before sharing a statement.

Try it — a tiny approval flow

Type this on the left:

flowchart TD
  A[Application] --> B{Docs complete?}
  B -->|Yes| C[Approve]
  B -->|No| D[Ask for docs]

A neat boxes-and-arrows
flowchart: Application →
a Yes/No decision →
Approve or Ask for docs.

Prefer clicking to typing? Switch to Draw mode and drag the boxes instead. When it looks right, export as PNG / SVG / PDF.

Try it — write once, paste anywhere pretty

## Agenda
- Disbursement update
- **Pending** KYC cases
- Next steps

A bold "Agenda" heading
with a clean bulleted list,
ready to paste into Teams
or an email — formatting intact.

Great for meeting notes, minutes and updates when you want headings and bullets without fighting a formatting toolbar.

When you'd use it

• You want to sketch an idea quickly and share a picture of it.
• You'd otherwise open a paid tool like Miro just for a rough diagram.
• You're explaining a flow to someone and want to draw it out live.

Your drawing is saved on your own device automatically, and nothing is uploaded.

Try it — a column into a query list

101
102
102
105

'101','102','105'

Handy when a report or a colleague hands you a column and a system (or a database IN (…) query) wants it on a single line. Toggle the quotes, separator, dedupe and sorting to taste.

Try it — turn a small CSV into JSON

name,city
Asha,"Pune, MH"
Ravi,Delhi

[
  { "name": "Asha", "city": "Pune, MH" },
  { "name": "Ravi", "city": "Delhi" }
]

1. Open the tool and pick CSV → JSON.
2. Paste your CSV. Keep First row is header ticked — it tells the tool that "name" and "city" are the column titles.
3. The JSON appears instantly. Notice it kept "Pune, MH" together as one value, even with a comma inside.
4. Click Copy or Download.

Going the other way (JSON → CSV) works the same. The Table viewer just shows a CSV as a grid with buttons to remove duplicate and empty rows.

Try it — clean up a messy list

Pune
Delhi

Mumbai
pune
Delhi

Delhi
Mumbai
Pune

1. Open the tool and choose Sort & dedupe lines.
2. Paste your list.
3. Tick Remove empty lines and Remove duplicates (with ignore capitals so "pune" and "Pune" count as the same), then click Sort A→Z.
4. Click Copy.

Other tabs: change case, clean & trim spacing, find & replace across the whole text, and a live word / line / character count.

Encode / decode — the same words, written differently

Base64, URL and Hex are just different alphabets for writing the same text, a bit like Morse code — used so text travels safely through systems that would otherwise garble spaces or symbols. You'll often be handed a jumble like UHJvdGl1bSDigrk= and want to know what it says.

UHJvdGl1bSDigrk=

Protium ₹

Hash — a fingerprint for text or a file

A hash is a short fingerprint of some text. The same input always gives the same fingerprint, and a one-letter change gives a completely different one — so it's how you check nothing was altered. It only goes one way (you can't turn a fingerprint back into the original).

abc

ba7816bf8f01cfea414140de5dae2223
b00361a396177a9cb410ff61f20015ad

JWT — reading the "digital wristband" after a login

When you log into a system it often hands your browser a long token called a JWT — think of it as a festival wristband that quietly says who you are and when it expires. This tool reads what's printed on the wristband: your name, when it was issued, when it runs out. It only reads the token — it does not check whether the wristband is genuine (that needs a secret key the system keeps).

1. Open the tool and go to JWT (an example is already filled in).
2. Paste your token — it splits into a header and a payload, shown neatly.
3. Read the expiry in plain time, e.g. "expired 2 hours ago" or "expires in 3 days" — flagged red if it's already dead. Useful when a system keeps logging you out and you want to know if your token simply ran out.

The ideas, in plain words

Control vs variant

The control (A) is what you do today. The variant (B) is the new thing you're trying. You compare B against A.

Conversion rate

The share of people who did the thing you care about — replied, clicked, completed. If 130 of 1,000 replied, that's a 13% conversion rate.

"Real or luck?"

Flip a fair coin 10 times and you won't always get 5 heads. Small groups wobble by chance. The statistics tell you whether a gap between A and B is bigger than that normal wobble.

Sample size

How many people you need in each group before the result can be trusted. Smaller improvements need far more people to prove — spotting a jump from 10% to 10.2% takes a huge audience; 10% to 20% needs very few.

Statistical significance & the p-value

The p-value is the chance you'd see a gap this big purely by luck if the two versions were actually the same. A small p-value means "luck is an unlikely explanation — the difference is probably real." The common bar is 5%: below that, we call it significant and trust it.

Confidence interval

Instead of a single tidy number, the honest answer is a range: "about 13%, somewhere between 11% and 15%." That range is the confidence interval — narrower when you have more data.

Statistical power

Your test's ability to actually notice a real difference when one exists. We usually aim for 80% power — meaning if B really is better, the test has an 80% chance of catching it. Too few people = weak power = you miss real wins.

The peeking trap

If you keep checking a running test and stop the moment it looks good, you'll fool yourself — random wobble alone will eventually look like a "winner." Either wait for your planned sample size, or use the peeking tool's stricter bar.

1. Before you start — "How many people do I need?"

So you don't run a test that's too small to prove anything.

Right now 10 out of 100
respond (10%). I want to
catch it if the new version
lifts that to 11%.

You need about
14,750 people
in each group.

2. After the test — "Did B really beat A, or was it luck?"

Type how many each version reached and how many responded; get a clear verdict.

A: 1,000 sent, 100 responded
B: 1,000 sent, 130 responded

B wins. Only about a 3.5%
chance this gap is luck —
trust the result.

That "3.5% chance it's luck" is the p-value. Below ~5% is the usual bar for "real". If it were 40%, the tool would say "not enough evidence yet — keep going."

3 & 4. The true rate, and peeking safely

• Confidence interval — the real response rate give or take, e.g. "13%, between 11% and 15%." Good for honest reporting instead of a single tidy-looking number.
• Peeking — if you plan to check a running test more than once, this gives you the stricter bar to use so repeated checking doesn't trick you into false winners.