Sudhanshu Gautam · Lead UX Designer, Wikimedia Foundation

I make complex things simple for people.

At the Wikimedia Foundation, I design how volunteers find what to translate in Content Translation, the tool behind over two million Wikipedia articles. I'm also designing how first-time internet users and newer editors create articles on mobile, one of the hardest tasks on Wikipedia, through Article Guidance. Before that: Wikipedia for first-time internet users on feature phones, and care tools at Noora Health, where I grew from design intern to Design Lead.

I start with the people problem, and I prototype in code: I was a software engineer first. Alongside my job, I've been designing and shipping my own AI side projects.

Read the case studies Download CV (PDF)

My life, so far

Case studies · Wikimedia Foundation

Designing for the people who read and build Wikipedia

I've worked at the Wikimedia Foundation since February 2020. I started on the Inuka team, in the Readers group, bringing Wikipedia to first-time internet users. In July 2024 I moved to Language and Product Localization, which makes the tools volunteers use to translate and write articles, and I'm still there. Each story says what I did, what the team did, and where the numbers come from.

Wikimedia Foundation · 2025 – 2026 · Lead UX Designer

Article Guidance

Demo video on Wikimedia Commons by Pginer-WMF, CC BY 4.0, trimmed and sped up

New editors' first articles are often deleted for missing sources, notability problems and the wrong structure, and patrollers carry the cleanup. I designed the full experience for the first experiment: title entry with Wikidata topic matching, source checks against community lists, a notability warning that offers other ways to help, community tips, and a starting outline in the Visual Editor, built mobile-first. An A/B test on red links in Simple English, Turkish and French Wikipedia compared 30-day survival. Guided articles survived at a 12% higher relative rate, statistically significant, and 30% higher on Turkish. Junior editors hit more friction, so the next phase cuts drop-off.

  • 12%higher 30-day survival
  • 30%higher on Turkish Wikipedia
  • 3pilot wikis in the A/B test
Project page ↗Experiment results ↗

Language and Product Localization · 2024 – now

Content Translation

Content Translation's Adjust suggestions dialog on Hindi Wikipedia, on a desktop and on a phone, with tabs for Collections, Regions and Topics

Recordings on Wikimedia Commons by UOzurumba (WMF): How to add a Featured collection and Featured collection, CC BY-SA 4.0, cropped and sped up

Content Translation already existed: volunteers use it to translate Wikipedia articles from one language into another. I designed a way for event organisers, the community members who run translation events, to make lists of articles for people to translate. I also introduced collections, which put those lists into translation suggestions, so any translator can find and use them, not only people at the event. A community can now feature one collection, and it becomes the default suggestion for everyone translating into that language.

  • 6Wikipedias featuring a collection
  • 2M+articles made with the tool since 2015
Live on every WikipediaProject page ↗
Read the case studyShow less

Content Translation was built before I joined the team. A volunteer takes an article in one language and publishes it in another, with machine translation to help. Often the hard part isn't translating. It's knowing what to translate.

Event organisers told the team they were “heavily burdened with finding actionable tasks”, and translators were getting “random and generic suggestions”. Since July 2024, when I moved from Inuka to the Language team, I've designed two things for this. Organisers can make a list of articles for their event. Collections then show those lists in translation suggestions, where any translator can pick from them.

2M+Articles created with Content Translation since 2015: the whole team's work, over years
AllWikipedias can use it, whatever the language
6Wikipedias had nominated a featured collection by 30 December 2025. The goal was five
Role

My part, and the team's

I'm one of the team's two designers, with Pau Giner, who is also its product manager. I write the design specs on Phabricator, design tests with translators, and do design sign-off on what ships.

What I did
  • Specified translation suggestions: topic filters, community lists that became Collections, sub-collections, and the tabbed “Adjust suggestions” dialog, built from Codex components.
  • Designed a self-guided prototype test for Wikimania 2024, and reworked the flow after an August 2024 usability round.
  • Designed Featured collections, from the banner to the badge, and signed off each build.
  • Designed dashboard details, like a “New contribution” button and a “Published as personal draft” label.
  • Wrote the design spec for each step of Article Guidance, a guided way for newer editors to create articles.
Who I worked with
  • Pau Giner, Principal UX Designer and Product Manager: base specs and design review. Section Translation and the original dashboard are Pau's designs.
  • Gerard Galofré, Senior Product Manager, and Purity Waigi, who wrote the 2024 collection tasks and usability findings.
  • Engineering: Santhosh Thottingal, Abijeet Patro, Kartik Mistry, Stephane Bisson, Eduardo Medina, Huei Tan and Nik Gkountas, with Niklas Laxström as Engineering Manager.
Decisions

What changed, and why

Each of these is in a public spec on Phabricator, so the reasoning is on the record, in the words we used at the time.

The Adjust suggestions dialog on Hindi Wikipedia: a search box, then automatic filters, CEE Spring 2025 collections, regions and culture topics
01

Suggestions people can steer

Long lists of filters “can be overwhelming, difficult to navigate, and hard to scale”, as I wrote in the spec. So Adjust suggestions groups them into tabs: All, Collections, Regions and Topics. Big campaigns like CEE Spring nest into sub-collections, using the “/” page names communities already use.

Testing had shown that people “wanted the ability to search as opposed to filtering”. In 2025, search moved above the tabs, “to prioritize search functionality”.

Shipped June and September 2025 · T392753, T374695, T398154

Translation suggestions with a Community Priorities banner for Wikiforhumanrights in Nigeria 2025, from French Wikipedia
02

Let communities say what matters

Collections on the dashboard were “both far (several clicks away) and chaotic”. With a featured collection, a community nominates one list of articles it needs, and it becomes the default for anyone translating into that language, “based on the target language of the translation pair, not the wiki the user is currently on”.

By 30 December 2025, six Wikipedias had nominated one: Arabic, Min-nan Chinese, Swahili, Igbo, Hausa and Telugu. The goal was five.

I wrote the design details for all five UI tasks · Shipped November and December 2025

The Published list, where one translation is labelled Published as personal draft
03

Say what state a translation is in

Translators “cannot easily identify which translations are still drafts”, so the list now says so, right under the title: “Published as personal draft”.

Shipped August 2025 · T395782

Article Guidance step one on a phone: write the article title, then pick the matching topicA New article page explaining that page titles can't include a bracket, and to remove it and try again
04

Article Guidance: guide new articles, don't rewrite them

Newer editors' first articles are often deleted for missing sources or the wrong structure. Article Guidance walks them through it: match the title to a topic, check sources against the community's own lists, read tips from experienced editors, then start from an outline. When a title can't be used, the screen explains why. As the spec says: “Do not rewrite what they typed.”

In an A/B test on pilot Wikipedias, guided articles showed a 12% relative improvement in 30-day survival, a statistically significant result (August 2026). Not everything moved: junior editors' gains weren't significant, and French Wikipedia saw little difference.

I wrote the spec for each step, with Pau Giner and the team · Live since July 2026

Beyond the team

Work other teams use

Some of this reaches past Content Translation, into the design system every Wikipedia shares.

Design system

A component in Codex

I designed Codex's ProgressIndicator: the Figma spec, the guidelines and the spinner. “ProgressIndicator was included in Codex v1.21.1. Thank you @SGautam_WMF for your contribution!”

Talks

Sharing the work

I co-presented the Foundation's latest features and tools at the Wikimedia Technology Summit 2024, with Niharika Kohli.

Sources
  1. Content translation and Translation suggestions, on mediawiki.org
  2. Featured collection, with the six Wikipedias
  3. Article Guidance updates, with the August 2026 results
  4. Specs on Phabricator: T392753, T398154, T408818, T395782, T432111, T416512, and Codex: T345921
  5. July 2024 newsletter, on the two teams merging

Inuka team · 2021 – 2024

Wikistories

The Wikistories logo and three phones: picking text from the Tiger article, a story about Borobudur, and a Read more page

Screen recording on Wikimedia Commons by Bennylin, CC BY-SA 4.0, trimmed and sped up

Short, visual stories made from Wikipedia articles and freely licensed images, piloted with the Indonesian Wikipedia community. I was the team's designer.

  • 3,738stories created
  • 261people made one
  • 38design tasks I wrote
Beta on Indonesian WikipediaProject page ↗
Read the case studyShow less

There was no easy way “to create or curate a visual narrative from Wikimedia content in a short snackable format for mobile devices”.

Wikistories was the Inuka team's answer: pick an article, choose freely licensed images and sentences from it, and publish a story of up to 10 pages. Some editors in the concept test said they never shared articles they edited, because “Wikipedia articles use formal language”. They said they “would definitely share the topics they edited through Wikistories”.

3,738Stories created on Indonesian Wikipedia since 2022, by 261 people
2,148Of them from one campaign, 14 to 24 September 2023, by 35 people
10Pages at most per story, about six seconds each
38Phabricator tasks I wrote, mostly design specs with mockups
Role

My part, and the team's

I was the only designer on the Inuka team. Wikimedia Indonesia ran the workshops and partnerships on the ground, and researchers ran the studies with editors in Indonesia and in Africa.

What I did
  • Wrote the design specs, with mockups, in 38 tasks from October 2021 to May 2024: the story builder and viewer, text and image selection, attribution, pacing, publishing and sharing.
  • Collaborated on the concept testing in Indonesia, and on the research in Kenya, Nigeria and South Africa.
  • Designed the Wikistories logo and wordmark for the pilot campaign.
Who I worked with
  • Angie Muigai, who led the first explorations in 2021, and Purity Waigi: Lead Product Managers.
  • Stephane Bisson, Tech Lead, with Eduardo Medina and Huei Tan; Connie Chen, data science.
  • Research: Eli Asikin-Garmager and Claudia Lo, with Ari Natarina, FloLab and Qhala.
  • Wikimedia Indonesia, who ran every workshop and GLAM partnership, and Bonaventura Aditya Perdana, Product Ambassador.
Decisions

What we chose, and why

The quotes are from the research reports and from my specs on Phabricator.

The Museum Nasional Indonesia article on Indonesian Wikipedia, with a round Buat Wikistories button just under the title
01

Start where people already tap

Stories sit as circles just under the article title. In concept testing, “all participants immediately understood the function of the image and (+) buttons”, thanks to “their familiarity in creating WhatsApp or Instagram stories”. I pushed to keep the create button on the left, near the title: “In absence of stories this look far from article title and might get less attention from users.”

T296725, T308247 · Screenshot of Indonesian Wikipedia by Bennylin, CC BY-SA 4.0

Mockup: a sentence about tigers selected in the article, with Highlight and ClearMockup: the highlighted sentence, with a row of tiger images to choose from
02

The article's own words, picked with a tap

People preferred picking text from Wikipedia over writing their own, “because it was faster, and the facts come from reliable sources”. So a story starts from the article. I designed text selection in 2021, then made it a single tap in 2023, where “the entire statement containing the tapped word/phrase is highlighted”.

T292773, T328906 · My mockups, October 2021

A story page about Borobudur, with the photographer's name and licence icons along the bottomThe Publish Wikistory screen, with a story title field and the licence terms
03

Only free images, always credited

Images come only from Commons. In the image search spec I wrote: “skip images that are hosted on Wikipedia as in many cases they are fair use images”, and anything under 640 by 640 pixels. The viewer shows each image's attribution “the same way it is shown in the StoryBuilder”.

T297130, T306504

Four mockup frames of a story about tigers: a cover image, the publish form, a title page and a page with text over the photo
04

Short, and paced for reading

Stories stop at 10 pages; in research, people used “around 5 images”. I proposed 7 to 10 seconds a page, since “they can always tap to advance faster, or long-press to pause”, and the team settled on 6. The publish button stays on screen, because “having a publish button available all the time make it easier for users to understand what to expect”.

T296494, T295982, T305116

A story with a share button in the viewerA Read more page listing three more stories about Borobudur
05

Made to be passed on

Sharing was a community request. I wrote the design for sharing a story from the viewer, and for a “Read more” page that leads into other stories. Sharing shipped in June 2023.

T303022, T310329

What happened

A pilot that stayed a pilot

Wikistories is still a beta feature on Indonesian Wikipedia, and it never expanded to other wikis. Most stories came from organised campaigns, and use is minimal now; the team removed its analytics in October 2025.

What carried on

The idea, in other projects

Wiki-Highlights, a later test project, continued the concept that started with Wikistories.

Sources
  1. Wikistories for Wikipedia and its FAQ, on mediawiki.org
  2. Wikistories Indonesia concept testing, on Meta
  3. Inuka team (archive), with the team list
  4. The Story creation log on Indonesian Wikipedia, read on 27 October 2026
  5. Specs on Phabricator: T292773, T297130, T295982, T303022, T328906

Inuka team · 2020 – 2021

Wikipedia for KaiOS

Three keypad phones running Wikipedia for KaiOS: the Tips list, the Tiger article, and the article text size slider

Introducing Wikipedia for KaiOS on Wikimedia Commons by Victor Grigas, CC BY-SA 4.0, trimmed and sped up

Wikipedia on a keypad phone, for first-time internet users, most of them on JioPhones in India. I took it through launch and a redesign shaped by the people using it.

  • 2.5M+downloads by Sept 2021
  • 20people in our usability study
  • 17%longer sessions, trending list
Launched 2020Try it in the simulator ↗
Read the case studyShow less

On a 2.8-inch screen with no touch, every action is a key press, and reading a long article means paging through it.

Most KaiOS owners were first-time internet users, “far less likely to know that Wikipedia exists—let alone use it on a regular basis”. I joined the Inuka team in February 2020, after Carolyn Li-Madeo had designed the app's early experience and the team had done its foundational research with 25 people in India. My job was to get it ready for launch, then improve it with what we learned from the people using it.

2.5M+Downloads since the first release, by September 2021
20People in the December 2020 usability study, in English, Hindi, Marathi and Telugu
107,600Readers in the trending-list experiment
+17%Longer sessions: 2 min 7 s up to 2 min 28 s
In motion

The app as it shipped

Recorded in the app's public simulator: one onboarding screen, Tips, search with the matched word in bold, the cue that there's more below, and the text-size slider.

The released app, from wikimedia.github.io/wikipedia-kaios, shown in KaiOS's own font, Open Sans.

Role

My part, and the team's

I was the designer for the launch and the next releases, working with a product manager, a tech lead and four engineers.

What I did
  • Designed the launch updates: onboarding, the About screen, the app icon and the store banners.
  • Tested on real phones, JioPhone 1 and 2 and the Nokia 8110 “banana” phone, and designed for their keys.
  • Worked on the December 2020 usability study with Daisy Chen and the research firm Hureo, then redesigned onboarding, text size, search, the gallery and settings from what we learned.
  • Designed the trending list: mockups, a D-pad prototype and loading states.
  • Wrote about the work, and co-presented it at Wikimeet India 2021.
Who I worked with
  • Angie Muigai, Product Manager.
  • Stephane Bisson, Tech Lead, with Gabriel Pita, Chelsea Riley, Eduardo Medina and Huei Tan.
  • Carolyn Li-Madeo, who designed the app's early experience before I joined.
  • Daisy Chen, Senior UX Researcher, and Hureo, the research firm in India.
Decisions

What the research changed

Most of these came straight out of the usability study. The phones below show the released app.

Onboarding: About Wikipedia, with a Get started button
Tips: How to read an article, How to search, How to switch between links, What is Wikipedia
01

One screen, then help when you need it

Most people in the study, 16 of 20, didn't read the onboarding descriptions: “users speed through the onboarding screens without reading through the content”. So I cut onboarding to one screen and moved the how-tos into Tips, which you open when you need them, each with a Try button.

Shipped 2021 · T270469, T270468

The Article text size slider over the Tiger article
The Tiger article in larger text
02

A text-size control you can see working

Text size used the 4, 5 and 6 keys, and some people “struggled to understand how to use it when they started pressing the keys(4,5,6) while the text size window was still open”. I replaced it with a slider that changes the article text live, right behind it.

Shipped 2021 · T267293

My mockups of changing the app language, with Hindi, Punjabi and Kannada at the top of the list
03

Indian languages first

Research found people didn't know Wikipedia was available in their own languages. So for JioPhone users, Indian languages come first in the language lists, matching the order of the phone's own settings. In the study, people “expressed that having Indian languages on top would help them within the article menu”, so that followed too.

T247255, T267290 · My mockups, 2020, CC BY-SA 4.0

The Tiger article's first page, with a small arrow at the bottom showing there's more below
04

Designed for the keys people actually have

I tested on JioPhones and a Nokia 8110, and that's where the small fixes came from: a cue that tells you there's more below the first page, confirmation kept on the right softkey “to avoid the accidental press”, and a flag that the app didn't support the JioPhone 2's extra key.

T244850, T262249, T247318

My mockups of a trending article list under the search box, with Cat, WWE and Cricket
05

Something to read, beyond search

People don't always open Wikipedia to look something up. I designed a list of articles trending in the reader's region. The team tested it with about 107,600 English-language readers in Nigeria, Pakistan, Tanzania and Uganda, and sessions grew from 2 min 7 s to 2 min 28 s.

The lesson the team took away: “Emerging market readers respond to locally-relevant reading suggestions.”

T271324, T271904 · Experiment results presented in 2021

Design system

Two icons in Codex

I drew the “network” and “networkOff” icons. They were merged into Wikimedia's OOUI icon set in December 2020, and Codex still ships them.

Writing

Sharing what we learned

I wrote “Improving the reading experience for smart feature phones” for Wikimedia's design blog in 2021.

Talks

Wikimeet India 2021

I co-presented the app with Angie Muigai at Wikimeet India in February 2021.

Sources
  1. Wikipedia for KaiOS, on mediawiki.org
  2. My article, with the 2.5 million downloads, September 2021
  3. Homepage content suggestions in the KaiOS app, the 2021 experiment results
  4. Specs on Phabricator: T270469, T267293, T267290, T244850, T271904
A 2018 Noora Health WhatsApp chat on a phone: two short Kannada audio messages around an illustration of a mother with a plate of nutritious food, made for families on limited data

Noora Health · 2016 – Jan 2020 · Design intern to Design Lead

Caregiver support after hospital

Families leave hospital trained by Noora's Care Companion Program, but recovery happens at home, alone. I did field research, co-designed the IVR voice-call pilot and the WhatsApp message system, and led the design team. In a 2017 trial with 293 families, 93% opted in and 83% still picked up the last message, but behaviour didn't change, so we moved to two-way WhatsApp. The programme I helped shape eventually grew to 2M+ subscribers across 4 countries and 11 languages.

  • 293families in the 2017 pilot
  • 93%opted in
  • 2M+subscribers by 2025
Annual Report 2025 ↗WHO Bulletin RCT ↗

Design tools for Wikimedia

Figma plugins for multilingual, open-content design

Wikimedia Commons Figma Plugin×

Wikimedia Commons Figma Plugin

Search Commons for freely licensed images and drop them into your designs, or fetch Wikipedia article content with its formatting intact: bold, italic, headings, links, superscripts and subscripts.

TypeScript, Figma Plugin API, Wikimedia REST API, Tailwind CSS · A live search, recorded in the plugin's interface

I18n Text Preview×

i18n Text Preview

Test UI layouts in other languages before shipping. Select a frame, pick a language, and the plugin translates the text in place with AI. Smart filters skip buttons, URLs, dates, placeholders and component instances.

Figma Plugin API, Claude API, local proxy server · The plugin's interface, with its translation server running

Side projects, vibe-coded with AI

Side projects, vibe-coded with AI

Personal projects I've built in my own time with AI coding tools, separate from my professional design work above. Open any of them for the full story.

Three Promptbook 1.2 screens: a message ready to send to ChatGPT, a note that ChatGPT's app didn't open so its website did, and the Exactly what goes sheet

iPhone app · 2026

Promptbook

Talk to any AI. Skip the typing. You speak, your iPhone writes it down, and one tap hands it to ChatGPT, Claude, Gemini, Grok or Perplexity.

  • 0servers or accounts
  • Youpress Send, never the app
  • 103unit tests in 1.2
1.0 on the App Store1.2 in TestFlightApp Store ↗
Read the case studyShow less

Talking to an AI on a phone means typing long prompts with your thumbs, and explaining who you are all over again to every app.

The AI apps do remember things, but their memory is written by the model, and in my research many people turn it off because they don't trust what it adds. So Promptbook takes the opposite approach: memory off, context on. You see and choose exactly what goes. Solo: research, product, design and the Swift build, with Claude Code.

0Servers or accounts. Your voice is turned into words on the iPhone
0Messages it sends for you. You tap send in the AI's own app
2Earlier notes it offers, at most, and only if you tap Add
25 daysFrom the first commit to the App Store, for 1.0
Five decisions

Designing for trust

Each one answers a question people have about AI tools. Where do my words go? Who presses send? What happens when it breaks? What does it remember? Can I take it back?

Welcome screen: Talk to any AI. Skip the typing. No account, no server. Your voice is turned into words on this iPhone.
01

Nothing leaves the phone unless you send it

No account and no server. Speech becomes words on the iPhone, and the build fails if the app's code ever references a network client.

A message in a speech bubble, Copied, To ChatGPT, and a Send to ChatGPT button
02

You press send, never the app

Send opens the AI's own app with your words in its message box, where that app accepts them. You tap send there. Promptbook never presses Send inside another app.

ChatGPT's app didn't open, so its website did. Paste your words there. Keep ChatGPT's answer? Paste.
03

Say what actually happened

When something misses, the screen says so in one line: “ChatGPT's app didn't open, so its website did. Paste your words there.”

The Exactly what goes sheet, showing the pinned context, an earlier dated note with ChatGPT's answer, and the request
04

Memory you can see and veto

An earlier note about the same person or thing is offered, dated, and only goes in if you tap Add. “Exactly what goes” shows the whole message first, and one switch stops the offers.

Remove this answer? Your words stay in your book. Keep it, or Remove.
05

Nothing is lost by accident

Redo keeps your first take until the new one is written down. Removing a kept answer asks first, and a book that can't be read is never overwritten.

Simulator screens from TestFlight build 1.2.0 (23), with a made-up message.

Failure states

Every Send has a way down

An iPhone app only opens from the links it claims, and each AI app claims different ones. So Send tries a short ladder of links. A website always opens bare, so your words never land in a browser's history. Words too long for a link go by paste, and the button says so.

  1. 1The AI's app, with your words in its boxFor the apps that take words from a link
  2. 2The app's own linkIf the first link misses
  3. 3Just the appThen you paste
  4. 4Its website, opened bareNo words in the link
  5. At every step, your words are already on the clipboard.

Until I've seen an app take the words on a real iPhone, the screen keeps a hedge: “If its box is empty, paste.” Gemini takes no words from other apps, so its button says “Open Gemini to paste”.

Simplifying

One decision after you speak

In build 21, the screen after recording had 16 controls, and I called it “a cognitive nightmare”. Build 22 cut it to one decision: which AI gets your words. In a quick check with five AI role-play participants working over real screenshots, “too many choices” fell from 6.4 to 3.6 out of 10. That's directional, not a study with real people, which still has to happen on a phone.

Build 21's busy sheet with many options, scrolled, next to build 22's simpler screens with one row of AI choices

Build 21 on the left, build 22 on the right, from the 1.2 design notes.

Left out, on purpose

No AI rewrite on the phone

An on-device rewrite can't run on an iPhone 13, couldn't be tested here, and Apple's model refuses harmless text. Instead, the background you pin goes in front of your words word for word.

How I'll know

A 30-day test, with a stop rule

Memory is only worth building if people use it: at least 4 in 10 offered notes added, and 1 in 5 week-four asks going out with an earlier note. If fewer than 2 in 10 offers are added, I stop and rethink, and Promptbook competes as a fast, private voice remote instead.

1.0.1 has been on the App Store since October 2026. 1.2 is in TestFlight, and which AI apps accept the words still has to be checked on a real iPhone.

Lenscord on a Mac and an iPhone: the menu bar panel, a video call and the iPhone pairing screen

iPhone + Mac apps · 2026

Lenscord

Your iPhone, as your Mac's camera. Plug it in with one cable and look great in Zoom, Meet, FaceTime and OBS.

  • 4Kinto OBS
  • 1USB cable
  • 0accounts or cloud
In private testinglenscord.com ↗
Read the full storyShow less

Mac webcams make people look worse than they really do. Lenscord turns the iPhone you already own into your Mac's camera, plus the tools to record like a pro.

It's two apps. The iPhone app captures and encodes the video. The Mac menu bar app hands it to video calls as a camera called “Lenscord”, and to OBS as a 4K source. The live video only travels over the cable to your own Mac. Solo: product, design and build, with Swift and Claude Code. iPhone build 27 went to TestFlight on 25 October 2026.

4KNative 3840 × 2160 into OBS, at 24 to 60 fps
1USB cable, with no Wi-Fi to drop mid-recording
0Accounts, cloud services or OBS plugins
10Test suites the release script runs before uploading a build
On the Mac

A calm menu bar app

The panel shows one status word, at most one button and three switches. Everything else lives in Settings. Setup goes Connect, Pair, Choose and Finish, and each step moves on by itself as soon as it works.

The panel

One status word, three switches, then the tools.

Lenscord menu bar panel: Live in an app, switches for Background blur, Darken background and Backup

Pair once

Type the iPhone's six-digit code once. After that, the two connect by themselves.

Setup step two of four: Type your pairing code

Choose where it goes

Video calls and other apps, OBS, or both.

Setup step three of four: Choose Lenscord in your app

A teleprompter that follows your voice

It scrolls as you talk. Speech is recognised on the Mac, never on a server.

Teleprompter window showing a script line in large white text

Look

Background blur and Darken background run on the iPhone.

Look settings: Background blur with Subtle, Medium and Strong, and Darken background

Mac screens from lenscord.com, with demo data. The product shot above is built from Lenscord's own marketing art: the person is an illustration.

On the iPhone

A camera app, not a control panel

One camera screen with two modes, Mac camera and Shorts, above a shutter row that never moves, like Apple's Camera. Opening it never starts the camera. In Shorts, the script sits right under the lens and waits when you pause.

Home screen: Your iPhone, as your Mac's camera, with the Mac camera and Shorts switch
Two modes
Pairing screen: Type this code on your Mac, 482 913
Pairing code
Shorts recording: the script in large text near the top and a red stop button
Shorts teleprompter
Videos list with Shorts and backups
Videos

iPhone screens from the current build and from lenscord.com, with demo data.

Language

One plain vocabulary

The iPhone, the Mac, the website and the docs share one small set of words. Status is always one short phrase from the same list, and some words never appear on screen at all.

ReadyLivePausedReconnecting…Not connectedRestarting the camera…
relayviewertakesslotphase
01

Don't drop the picture

If the cable wiggles or a call comes in, OBS keeps the last picture and carries on in the same recording. Backups are written in parts, so even a crash leaves a playable file.

02

Say what's happening

When the phone gets hot, Lenscord turns effects off first and lowers the frame rate only as a last resort, never the picture size, and both apps say so in one line.

03

Honest about the alternatives

Apple's Continuity Camera is free and built in, so the site compares all three options with sources: “For quick calls, it's great. … Lenscord is for when you're making something.”

Vedit Studio on a Mac, with two reel covers it made

Video editor · 2026

Vedit

Raw footage in. A finished short out. The editor I built to turn my recordings into YouTube videos and Shorts.

  • 9:57from 11:22 raw
  • 50synthesised sounds
  • 0LLMs in Autopilot
Personal toolWatch me use it ↗
Read the full storyShow less

I record in Hinglish and English: a 4K talking head in OBS, clips from my iPhone, and screen recordings of the tools I'm testing.

Vedit turns that into a finished 16:9 YouTube edit plus vertical Shorts and Reels: cut, captioned, colour-graded, scored and checked. As a Claude Code skill, Claude makes the editorial calls. Vedit Studio, the Mac app, has an Autopilot that runs on rules, my choices and local models like Whisper, with no language model at all.

9:57The finished video about Vedit, cut from an 11:22 raw recording
50Sound effects it synthesises, with no stock music or samples
7Styles the Autopilot can render
0Language models inside the Autopilot
In motion

Real reels it made

Each style is a full visual system with its own captions, graphics and sound: word-pop captions and stickers, a terminal look, explainer cards that build up, and a race between three AIs with live timers.

Plum-pop
Terminal
Cards
One prompt, 3 AIs

Clips from reels Vedit rendered, muted here.

Pipeline

Every step, checked before export

It cuts retakes and fillers, writes Hinglish captions that pop with my voice, and hides emails and OTPs with masks that follow the screen. Then it grades, adds music and sound, and runs a QA check on lip sync, loudness, captions and privacy.

  1. 01Ingest
  2. 02Clean cut
  3. 03Captions
  4. 04Privacy
  5. 05Look
  6. 06Graphics
  7. 07Sound
  8. 08Render
  9. 09QA
  10. 10Deliver
A QA board of six vertical frames, marked with face boxes, safe areas and caption sizes

A QA board from a sample edit: faces, safe areas and text sizes marked on every beat.

Rule 01

Never touch the originals

Projects point to the footage, and the originals are never modified or uploaded.

Rule 02

Hide anything private, everywhere

Credentials, codes, personal details and other people's chats are masked for the whole video.

Rule 03

Accuracy beats hype

Every claim on screen has to be true, and visible or said in the footage.

Six video thumbnails and one Short from the channel

YouTube channel · since May 2026

Learning AI in public

A designer trying AI agents, design and coding tools on camera, in English and Hinglish: what worked, what didn't, and what I'm still figuring out.

  • 14videos
  • 40Shorts
  • 2languages
Active@sudhanshugtm ↗
See the videosShow less
Videos

Honest tests of new tools

“I'm not an AI expert. I'm learning by doing: building small things, testing new workflows, and finding useful ways to bring AI into work and life.” That's the channel's own description. Counts are from October 2026.

Gemini 4 Argon Review: Ek Photo Se 3D Game Ban Gaya (Hindi) AI meri job le lega? Ek product designer ke honest two cents Maine apni YouTube videos AI se edit karwayi | Claude Opus 5.5 Grok Build ne meri iOS App bana di! Build Fail se Working App tak Claude Fable 5 Just Redesigned My Real Website (Honest Test) Grok Build CLI: A Designer's First Look Codex CLI ke 5 naye updates: ab ye BAAT bhi karta hai Superfast Review of DeepSeek Harness on MacBook
Codex CLI ab BAAT karta hai AI ne meri YouTube videos edit ki Gemini 4 Argon ne mera living room 3D game bana diya $20 mein kaunsa AI lein? Screen pe email ya OTP? AI khud blur karta hai AI ne mere video ka music khud banaya

Website · 2026

whichcli.com

Does your software have a CLI? A verified directory of 78 developer tools, with install commands and the publisher's own docs.

  • 78tools
  • 0guessed commands
  • AI-runhuman-accountable
Livewhichcli.com ↗
Read the full storyShow less

When you work with coding agents, it helps to know whether a product has an official command-line tool, what the command is called and how to install it safely.

whichcli answers that for 78 developer products, from Claude Code and Stripe to Figma and Docker. Every answer links to where it came from, and the site looks and behaves like a terminal you can browse from the keyboard. A solo project; AI agents keep the records up to date, and I'm accountable for them.

01

Evidence, not guesses

Records show status, executable, install command and the date verified, with primary sources.

02

Statuses that say exactly what's true

Figma's CLI is marked “Official · scoped”, because it “is not a general Figma file or design CLI”.

03

Works without JavaScript

All 78 entries stay readable with JavaScript off. JavaScript only adds the interactive previews.

Website · 2025 – now

Indian Coffee Guide

India's specialty coffee, made simple. Three taps, no quiz: how you brew, a roast and a budget narrow about 960 coffees to a few.

  • 75roasters
  • 34cities
  • Dailyprice checks
Liveindiancoffeeguide.com ↗
Read the full storyShow less

I built Indian Coffee Guide to make it easy to discover Indian specialty roasters and their coffees in one place.

Product details come from the roasters' own websites. Prices are shown per bag and per 100 g, checked every day, with a list of what got cheaper. Every roaster has a page, and each coffee links straight to the roaster's shop. Built with Next.js, Supabase and Vercel; numbers from October 2026.

75Roasters in the guide
34Cities they roast in
~960Coffees, each with its price per 100 g
DailyPrice checks, with a list of what got cheaper

A page for every roaster

City, founding year, active coffees, roast spread and price range, with a link to shop direct.

Blue Tokai Coffee Roasters page on Indian Coffee Guide

On a phone

The same three-tap finder.

Indian Coffee Guide homepage on a phone

Skills and widgets

Small tools, made for how I work

design-side-view Open source

Keeps the agent's chat on the left and opens mockups in a pane on the right, inside Kitty. Works with Grok Build, Claude Code, Codex and Cursor.

GitHub →Full video →

Focus mode Claude Code

Start it with /timer 1h. It reads my Google Calendar to know what I'm working on, plays binaural focus audio, and checks in every 15 minutes.

Full video →
Vedit Studio

video-edit Claude Code

The skill behind Vedit. Claude edits my raw recordings into finished YouTube videos and Shorts, and only stops to ask about real blockers.

cartoon-builder In progress

Makes episodes of Pip, a toddler cartoon about a turquoise baby dinosaur and his snail friend: story, keyframes, voice, animation, music and QA. The visuals are AI-generated. A first 34-second sample is made.

phabricator Open source

Read-only access to Wikimedia Phabricator for coding agents: look up tasks, search, browse related tasks and view attachments.

GitHub →

Session Browser

Browse and resume past AI coding sessions from Claude Code, Codex, OpenCode, Copilot, Pi, Gemini and Grok, right from the bar.

GitHub →

Wiki Wait

A calm Wikipedia reader for while an agent works. Pick a mood and follow a good rabbit hole without losing sight of the work.

GitHub →

SQLite Viewer

Every SQLite database touched in the last 60 days, at a glance. Click one to open it in DB Browser.

GitHub →

Scheduled Jobs

What runs next on this machine, and whether the last run worked. systemd timers and cron, read-only.

GitHub →

Boot macOS

One click to reboot an Asahi Linux Mac into macOS, for the next boot only.

GitHub →

How I work

People first. The tools keep changing.

01Start with the people problem

I've designed for first-time internet users on feature phones, families continuing care after a hospital stay, and volunteer translators across hundreds of Wikipedias. The research comes first; the screens come later.

02Adapt to whatever ships

I'm comfortable in the terminal with Claude Code, Codex, Grok Build, Gemini CLI, Cursor, OpenCode, Copilot CLI and Pi, and I wire agents together when a job needs it. When a new model ships, I try it on real work.

03Explore by building

I'm really enjoying exploring AI. I build small things to learn, and some become real products: Promptbook went from first commit to the App Store in 25 days. I share what worked, and what didn't, on YouTube.

Comfortable withClaude CodeCodexGrok BuildGemini CLICursorOpenCodeCopilot CLIPiMCPAgent skillsFigma
Also triedAiderGooseAmpKimiQwen CodeDeepSeek Harness

Say hello

I'm happy to talk about designing for people, AI products, or anything on this page. I'm open to relocating to London or Dublin.

Sudhanshu9252@gmail.com Download CV (PDF)