Your markdown file is the source of truth. Substack is a render target.
## Ship a post in three commands
```bash
pipx install git+https://github.com/HighnessAtharva/substack-cli
substack init
substack push my-post.md
```
| **Push** A markdown file becomes a Substack draft. Headings, bold, links, code blocks, lists, quotes and captions all convert. | **Pull** A live post comes back down as markdown. Every image downloads next to it. | **Update** You rewrite an already published post in place. No email, no feed bump, same URL. |
| **Audit** You see exactly what an update would destroy, before it destroys anything. | **Publish** You go live now, with or without emailing your subscribers. | **Schedule** You set a real server-side release time, and you cancel it the same way. |
| **Notes** You post a Substack Note from a file or a string, with image attachments. | **Covers** The hero image uploads itself. `cover` swaps it on a live post and leaves the text alone. | **Slugs** Your file owns the public URL. Substack never gets to invent a truncated one. |
| **Tables** A markdown table renders to a PNG, because the Substack editor has no table support. | **Templates** You list the saved post templates on your account and push a file against one. | **Sitemap** You build a local index of every live post, ready to commit beside your drafts. |
## The problem in one paragraph
A number lands in paragraph nine of an article. Six months later the source changed, or
the sentence around it got edited, and nobody rechecked. No normal tool can see that. A
spell checker cannot. A linter cannot. A person rereading their own work cannot, because
the sentence still reads fine.
groundtruth makes the link explicit and machine-checkable. Next to each claim you record
the source it rests on, plus the exact words in that source that support it. The tool
refuses a claim it cannot find in your document. It refuses a quote it cannot find in the
source. Then it fails your build when either side moves.
**What a skeptic needs to hear:** this checks that a claim is *anchored*, not that a claim
is *true*. A wrong source you quoted correctly still passes. What you get is one page that
shows every sentence in your document resting on nothing.
---
## What you get
### One command in CI, and a real reason for the failure
## Install
```bash
pip install git+https://github.com/HighnessAtharva/quotecard.git
```
```bash
playwright install chromium
```
```bash
quotecard init --name "Your Name" --handle yourhandle --avatar photo.png
```
Chromium does the rendering and installs once, about 150 MB. `init` stores your identity, so plain text always comes out with your face on it. Then cache the fonts, which makes every later render offline and identical on any machine:
```bash
quotecard fonts
```
Check the whole setup in one command:
```bash
quotecard doctor
```
---
## Thirty seconds
```bash
quotecard https://x.com/username/status/1234567890
```
```bash
quotecard "the line I want on a card" --look editorial-cream --size 1:1
```
```bash
quotecard -f links.txt -d cards/ --look dark --size 4:5,9:16
```
Cards land as 1080-pixel PNGs in the current directory. Send them elsewhere with `-d cards/`, or name one file with `-o card.png`. Add `--open` to pop the result straight into your image viewer.
---
## Eight looks
Run `quotecard looks --render` to make this sheet yourself, with your own words and your own face on it.
| | | | |
|:--:|:--:|:--:|:--:|
| 
| 
| 
|
|:--:|:--:|:--:|
| **`4:5`** 1080x1350
|
| `quotecard https://substack.com/@a16z/note/c-321561696 --look dark` | 
|
| `quotecard https://blog.example.com/p/a-post --look serif-light` | 
|
| `quotecard "writing is thinking." --format text --quote-mark`
|
More, without pictures:
```bash
# every look at once, then pick the winner
quotecard https://x.com/user/status/123 --all-looks -d cards/
# a story and a feed card from one link
quotecard https://x.com/user/status/123 --size 9:16,4:5 -d cards/
# fix a typo, drop the counts, add your site in the footer
quotecard https://x.com/user/status/123 \
--text "the corrected line" --no-engagement --watermark "yoursite.com"
# a long quote out of a file, newlines kept
quotecard placeholder --text-file quote.txt --format text --look serif-dark
# print-ready, 2160 pixels wide
quotecard "a line" --scale 2 --size 1:1
```
---
## Whose face goes on the card
Pick one of three modes in config, then override it per run.
| Mode | Behavior |
|---|---|
| `auto` (default) | your identity on your own text and on posts under one of your handles, the real author on everybody else's |
| `always-me` | your identity on every card |
| `source` | the real author every time |
```bash
quotecard init --alias yoursubstack --alias yourgithub
quotecard config --set identity=always-me
quotecard https://x.com/someone/status/123 --me
```
Attribution matters, so `auto` keeps a stranger's name and face on a stranger's words. When a source gives no photo, your picture fills the slot instead of leaving a hole.
---
## Batch
One link or one line of text per line. Blank lines and `#` comments are ignored.
```bash
quotecard -f links.txt -d cards/ --look dark --size 4:5,9:16 --json
```
Every link is read at the same time, a repeated link is read once, and one browser renders the whole file. Twelve cards from nine links take about six seconds. Tune the pool with `--jobs`, which defaults to 8.
Cards that already exist are skipped. Pass `--overwrite` to replace them.
---
## For agents and scripts
`--json` prints one object per card on stdout. Nothing else goes to stdout, so the stream stays parseable.
```json
{"input":"https://x.com/...","ok":true,"path":"cards/x-user-123.png","width":1080,"height":1350,"look":"light","font_size":57.6,"source":"x","author":"harsehaj","handle":"harsehaj","text":"...","url":"https://x.com/..."}
```
Exit codes: `0` every card worked, `1` nothing worked, `2` some worked and some failed. A failed link never stops the batch, so read `ok` per line.
```bash
quotecard fetch https://x.com/user/status/123 # read a post, render nothing
quotecard -f links.txt --dry-run --json # plan a batch, no browser
```
A card whose text hit the minimum size carries `"warning"`. Move that post to `--size 9:16` or split it, and never ship it clipped.
---
## Every option
```
quotecard [inputs...] [options]
-f, --from-file FILE one URL or line of text per line
-o, --out FILE exact output path, for a single card
-d, --out-dir DIR where the PNGs land
--look NAME comma separated, or --all-looks
--size SPEC comma separated, named or WxH
--text-size SIZE small, medium, large, xl
--format card|text card keeps the header, text drops it
--align top|center|bottom
--fit auto|shrink|fill
--scale N 2 renders at 2160 pixels wide
--pad PX canvas around the card, 0 fills the frame
--quote-mark open with a large quotation mark
--watermark TEXT small footer text on the right
--avatar-shape circle|square
--text TEXT replace the fetched text
--text-file FILE read the card text from a file
--title TEXT bold line above the body
--name, --handle, --avatar
--me force your identity onto every card
--identity MODE auto, always-me, source
--no-engagement hide the counts
--no-verified hide the badge
--likes N, --reposts N, --replies N
--jobs N parallel fetches, default 8
--offline block all network during render
--overwrite replace PNGs that already exist
--dry-run fetch and plan, render nothing
--open open each finished card
--json one JSON object per card
-q, --quiet
```
| Subcommand | What it does |
|---|---|
| `init` | store your name, handle and profile picture |
| `config` | show or change stored defaults |
| `fonts` | cache the webfonts for offline, identical renders |
| `looks` | list every look, or `--render` a sample of each |
| `fetch` | read a post and print it as JSON |
| `doctor` | check config, fonts, Chromium and the network |
---
## How it works
Four readers hand back one shape, and one Chromium renders it.
- **X** goes through the public syndication endpoint behind every embedded tweet. No key, no login.
- **Substack** notes come from the public reader endpoint, and articles from the publication's own posts API, custom domains included.
- **Threads** serves an empty page to a plain HTTP client, so the same Chromium that draws the card reads the post.
- **Text** skips the network.
The card is one HTML file. The text auto-fits by binary search, so a 2,000 character post shrinks until it fits and a six word line grows to fill a story.
Speed comes from four choices. Every link in a batch is read at once. One browser and one page serve the whole run. Avatars download once and are cached. The fonts live on disk, so the render never waits on a request.
---
## Config
`quotecard config --path` prints the file. `QUOTECARD_HOME` moves the whole directory.
| Key | Default | Meaning |
|---|---|---|
| `name`, `handle`, `avatar` | empty | your identity |
| `verified` | `false` | show a badge on your cards |
| `identity` | `auto` | auto, always-me, source |
| `aliases` | empty | other handles that are you |
| `look`, `size`, `text_size`, `format` | `light`, `4:5`, `large`, `card` | defaults for every run |
| `engagement` | `true` | show counts when the source has them |
| `out_dir` | `.` | where PNGs land |
```bash
quotecard config --set look=dark --set size=1:1
```
---
## Tests
```bash
pip install -e ".[dev]" && pytest
```
147 tests. Parsing runs off saved API payloads, so most of the suite needs no network. `pytest -m "not render"` skips the tests that start Chromium.
---
## Limits
- Threads needs Chromium to read a post, which is slower than the other three, and Threads gives no engagement counts.
- A protected, deleted, or age-gated X post cannot be read. The error says so.
- X's public endpoint carries no repost count, so an X card shows likes and replies.
- Images and video inside a post are not drawn. The card carries the words.
---
## License
The license is MIT. Built by [Atharva Shah](https://atharvashah.com).
---
# 9x9 Goals Grid: The Harada Method
## Overview
The **9x9 Goals Grid** is a digital implementation of the Harada Method, the same framework used by Shohei Ohtani to transform from a high school athlete into a baseball legend. Unlike traditional to-do lists that are infinite and overwhelming, this tool forces finite prioritization. It turns one abstract dream into **81 concrete, actionable steps**.
Designed with a **"Neobrutalist"** aesthetic and a **"Privacy-First"** architecture, it lives entirely in your browser. No logins, no cloud databases, no tracking. Just you and your goals.
## Video Tour
[Watch the full tour and implementation details](https://www.loom.com/share/05e45cb8aba14ef1b7a7fb980c560dae)
## The Philosophy
Productivity tools often become procrastination traps. We spend more time organizing lists than doing the work. The 9x9 Grid solves this by enforcing constraints:
- **Constraint 1: The Space is Finite.** You have exactly 8 subgoals and 64 actions. If it doesn't fit, it's not a priority.
- **Constraint 2: Visual Impact.** Seeing your entire roadmap in a single glance creates a powerful mental model of your path to success.
- **Constraint 3: Complete Privacy.** Your ambitions are personal. By using LocalStorage, we ensure your data never leaves your device.
## Core Features & Workflow
### 1. The Macro View (The Bird's Eye)
At the center sits your **Main Goal**. Surrounding it are 8 **Subgoals**, the pillars required to support that main goal (e.g., Mental Strength, Physical Health, Technical Skill).
Each of these pillars is further broken down into 8 specific, actionable tasks. This recursive structure ensures that even the loftiest ambition is grounded in daily habits.
### 2. The Focus Mode (The Deep Dive)
It can be daunting to look at 81 boxes at once. **Focus Mode** allows you to zoom in on a single 3x3 sector. This isolates one specific domain of your life, like "Finance" or "Health", allowing you to plan that area without distraction.
This mode is designed for deep work sessions where you are brainstorming the specific "How" for a particular "What".
### 3. The Narrative View (Chapters)
Not everyone thinks spatially. The **Chapters View** transforms the grid into a linear, card-based story. It presents your 8 subgoals as distinct chapters in the book of your success, perfect for reviewing your strategy or presenting it to mentors.
## Technical Architecture
This application was built with performance and portability in mind.
- **Framework**: React 18 with TypeScript for type safety and component modularity.
- **Build Tool**: Vite for lightning-fast development and optimized production builds.
- **Styling**: Tailwind CSS for the stark, high-contrast "Neobrutalist" design system.
- **State Standards**: LocalStorage API for persistent, client-side data management.
## Export & Durability
Because the tool is local-first, data portability is key. Users can export their entire board as a **JSON file** for backup or transfer between devices. Additionally, the tool allows generating high-resolution **PNG/SVG** images of the completed grid, perfect for printing as a poster or setting as a desktop wallpaper to keep your goals visible.
---
# AccuKnox Control Center
The **AccuKnox Control Center** is an internal hub that brings the marketing team's day-to-day tools under one roof. Instead of scattered spreadsheets, shared docs, and Slack links, the team has a single URL for everything operational.
## Five Apps in One Place
### AK Socials
Tracks the email and LinkedIn post schedule across campaigns. See what's going out, when, and who owns it, without digging through campaign calendars or asking in Slack.
### AK Spends
Monthly and annual marketing spend at a glance. Budgets by channel, actuals vs. targets, and running totals. Built for marketers, not finance teams, no pivot tables required.
### Login Helper
A secure way to share login instructions with teammates. Instead of sending credentials over Slack or email, Login Helper generates a structured, shareable card with the right context attached, environment, usage notes, and who to contact if something breaks.
### AK Pricing
Generates pricing quotes quickly from a predefined product catalogue and pricing matrix. Useful for sales support, partner inquiries, and internal approvals where a clean quote needs to go out fast.
### MKT Performance
Tracks team member goals and targets against actuals. Part OKR tracker, part accountability layer. Teams can see their numbers in real time without waiting for end-of-month reports.
## Why It Exists
AccuKnox's marketing team runs across campaigns, content, DevRel, and demand gen simultaneously. Every function was using a different tool to track its work. The Control Center consolidates these into one interface so onboarding is faster, context-switching is lower, and the team spends less time hunting for information.
The hub is internal. It is infrastructure for the team that ships the products, and it is not a product itself.
---
# Algorithm Listens
## Overview
Algorithm Listens is an ongoing computational text analysis series that explores artists through their complete bodies of work. Each essay combines lyrical analysis, sentiment mapping, and thematic clustering to provide deep insights into an artist's creative evolution and thematic patterns.
## Project Goals
- Analyze **100 artists** across **10 genres**
- Provide data-driven insights into lyrical themes and patterns
- Map sentiment and emotional trajectories across discographies
- Identify thematic clusters and recurring motifs
- Understand artistic evolution over time
## Methodology
### Data Collection
- Complete lyrical datasets for each artist
- Metadata including release dates, album information, and collaborations
### Analysis Techniques
- **Natural Language Processing (NLP)** for text analysis
- **Sentiment Analysis** to track emotional tone
- **Topic Modeling** to identify recurring themes
- **Cluster Analysis** to group similar songs and albums
- **Time Series Analysis** to track evolution
### Visualization
- Interactive charts and graphs
- Sentiment heatmaps
- Thematic networks
- Word clouds and frequency distributions
## Technology Stack
- **Python** for data processing and analysis
- **Pandas** for data manipulation
- **NLTK/spaCy** for natural language processing
- **Scikit-learn** for machine learning and clustering
- **Matplotlib/Plotly** for visualizations
## Artists Covered
This project aims to cover 10 artists from each of the following genres:
- Pop
- Rock
- Hip-Hop/Rap
- R&B/Soul
- Electronic
- Country
- Jazz
- Metal
- Indie/Alternative
- Folk
## Key Insights
Each artist analysis provides:
- **Lyrical vocabulary complexity** over time
- **Dominant themes** and how they evolve
- **Sentiment trends** across albums
- **Collaboration patterns** and their influence
- **Genre-specific characteristics** and innovations
## Future Development
- Expand to more artists and genres
- Add audio feature analysis
- Integrate streaming data for popularity correlation
- Create interactive web visualizations
- Build recommendation systems based on thematic similarity
---
_This project represents a deep dive into the intersection of data science, music, and literary analysis._
---
# BuyDip
BuyDip is a proactive investment tool designed to help investors make rational decisions by removing emotion from the process. The tool provides a dashboard that shows how far various market sectors and indices are down from their all-time highs. Based on a user's predefined investment strategy and custom thresholds, it recommends specific actions to take when the market dips. This allows investors to execute their plans with discipline, without the need for constant market monitoring or second-guessing.
**DEMO:** [https://www.atharvashah.com/buydip/](https://www.atharvashah.com/buydip/)
---
## Key Features
- **Emotion-Less Investing:** The tool provides clear, predefined actions based on market drops, helping users avoid panic and indecision.
- **Customizable Thresholds:** Users can set their own triggers, such as "down 5% = small buy" or "down 15% = all-in," tailored to their personal investment strategy.
- **Dual Modes:**
- **Simulation Mode:** Allows users to test different strategies and thresholds with a simple slider to see how their action items would trigger.
- **Live Mode:** Pulls real-time market data to show current drops and recommends actions based on the user's live portfolio and settings.
- **Automated Alerts:** The system actively monitors the market and provides alerts, so users don't have to constantly check charts.
- **Personalized Dashboards:** The tool can be customized for individual users with their preferred tickers, mutual funds, and unique investment variables, all without needing to write any code.
---
## Technical Details
- **Docker Containerization:** The application runs within a Docker container, making it easy to deploy and manage.
- **Configuration-Based Logic:** User-specific settings, such as investment thresholds and target tickers, are stored in a separate configuration file. This allows for easy customization without altering the core codebase.
- **Data Retrieval:** The tool pulls live market data from various sources to provide up-to-the-minute information on market performance.
- **User Authentication:** Access to the live dashboard is protected by a password, ensuring data security.
---
## Learnings
Building this tool provided valuable insights into creating a user-centric application that solves a real-world problem for investors. The project reinforced several key concepts:
- **Simplifying Complex Data:** The primary challenge was to take complex market data and present it in a simple, actionable format. This involved creating a clean dashboard that highlights what's most important to the user: the current market drop and the corresponding action to take.
- **User-Centric Design:** The inclusion of both a simulation mode and a live mode was crucial. The simulation mode helps build user confidence and allows for risk-free experimentation, while the live mode provides the practical utility of the tool.
- **Building a Customizable and Scalable Solution:** By separating user configurations from the application's core logic, the tool was made highly customizable. This "no-code" approach for the end-user makes it accessible to a wider audience and simplifies the process of setting up a new dashboard for each user.
- **Practical Application of Automation:** The project showcased how automation can be used to improve decision-making. By setting up the tool to actively monitor and alert based on triggers, investors can be more disciplined and less reactive.
---
# CaptionThis
**CaptionThis** is a daily 4-panel comic strip that uses sarcasm, wit, and irony to roast finance, tech, media, fashion, and urban cultural trends. It features two contrasting characters, Leo and Nina, who dissect modern life with deadpan humor.
## The Characters
- **Nina (The Cynic):** Sharp, pragmatic, and allergic to hype. She cuts through the noise with razor-sharp sarcasm.
- **Leo (The Optimist):** Enthusiastic, impulsive, and always chasing the next big thing. He sets them up; she knocks them down.
## AI-Driven Production Workflow
CaptionThis leverages a sophisticated AI pipeline to maintain a daily publishing cadence while ensuring character consistency and topical relevance.
### Step 1: Trend Discovery & Research
We utilize **Perplexity AI** to scan for high-engagement topics across finance, tech, and pop culture. The system filters for "roast-able" news items, stories that highlight hypocrisy, absurdity, or modern struggles.
### Step 2: Script Generation (LLM)
Selected topics are fed into an LLM with a strict **Persona-Based Meta-Prompt**.
- **Input:** Trending headline (e.g., "NFT market crash").
- **Processing:** The model applies the distinct voices of Leo and Nina. It enforces a 4-panel structure: Setup -> Reaction -> Escalation -> Punchline.
- **Output:** A structured script optimized for brevity and visual storytelling.
### Step 3: Structured Visual Prompting
To solve the common AI consistency problem, we don't just ask for "a comic strip." We use a **Master JSON Prompting Strategy**.
- **Style Enforcement:** A rigid JSON schema defines the art style (black & white, high contrast, minimalist linework) to prevent style drift between panels.
- **Character Consistency:** Detailed attribute dictionaries for Leo and Nina are injected into every prompt, ensuring their clothing, hair, and features remain constant.
- **Panel Composition:** The prompt specifies camera angles and speech bubble placement to ensure the text remains legible.
### Step 4: Generation & Assembly
The JSON prompts drive the image generation model to create individual panels. These are then assembled into the final 4-panel strip, ready for Instagram carousels and X threads.
## CaptionThis. Brand Snapshot
**Concept:** Daily 4-panel comic roasting finance, tech, media, fashion, and urban culture using sarcasm and irony.
**Tone:** Snarky. Relatable. Bold. Satirical.
**Style:** Minimal visuals. Text-first punchlines.
### Characters
- **Leo:** Optimistic, naïve, deadpan setup guy. Loves tech and trends.
- **Nina:** Cynical, sharp, sassy closer. Hates jargon and hype.
- **Dynamic:** Optimist vs cynic. Setup vs punchline.
### Audience
18 to 35. Gen Z and Millennials. Meme-native. Into money, tech, culture, and internet humor.
### Content
- **Format:** 4 horizontal panels.
- **Cadence:** Daily.
- **Types:** Trend roasts, evergreen urban struggles, caption contests, weekly best-of.
### Platforms
Instagram. Twitter/X. TikTok. Reddit.
Growth via daily posting, trend hijacking, meme page cross-shares.
---
# Criterion Collection Scraper
The problem statement revolves around the need for a web scraper designed to extract detailed information about movies from the Criterion Collection. The scraper is tasked with collecting various attributes such as movie title, director, media type, country, year, runtime, aspect ratio, language, Blu-ray and DVD availability, prices, poster URL, page URL, and thumbnail URL. The ultimate goal is to facilitate easy and programmable access to this movie data, catering to both developers and movie enthusiasts.
**GITHUB:** https://github.com/HighnessAtharva/Criterion-Collection/
## Learnings
Building this web scraper involves navigating through the Criterion Film Collection page, utilizing XPath to locate essential data points, and simulating clicks on individual film pages to gather more detailed information. The extraction process includes cleaning the acquired data to maintain a uniform format. The scraper then serially compiles this information and stores it in an Excel Workbook for easy reference.
## Pipeline Overview
1. **Visit Criterion Film Collection Page:** Start by navigating to the listing page (https://www.criterion.com/shop/browse/list) of the Criterion Film Collection.
2. **XPath Extraction:** Locate the XPath of key data items on the page, which will serve as identifiers for scraping.
3. **Simulate Clicks:** Simulate clicks on each film page to access more detailed information about the movies.
4. **Data Fields Extraction:** Extract relevant fields such as movie title, director, media type, country, year, runtime, aspect ratio, language, Blu-ray and DVD availability, prices, poster URL, page URL, and thumbnail URL.
5. **Data Cleaning:** Ensure the extracted data is in a uniform and clean format for further processing.
6. **Excel Workbook:** Serially add the collected movie information to an Excel Workbook, providing an organized structure for easy storage and retrieval.
## Future Enhancements
Looking ahead, the scraped data can be leveraged to create an API, offering a seamless way for developers and movie enthusiasts to access and interact with the Criterion Collection movie details programmatically.
This web scraper simplifies the process of gathering extensive movie information, making it a valuable tool for those who wish to explore and utilize Criterion Collection data effortlessly.
---
# DevRel Workbench
**DevRel Workbench** is a professional toolkit for developer advocates, a single place for playbooks, AI-assisted tools, curated resources, and a community layer. Built to help DevRels move faster and ship better work.
## Playbook Library
15 curated, step-by-step playbooks covering every major DevRel motion. Each playbook has a unique shareable URL so you can link a teammate directly to the right workflow.
Playbooks cover:
- **Product Launches.** pre-launch narrative building, developer outreach, launch day coordination, post-launch content sequencing
- **Community Building.** Discord/Slack setup, contributor onboarding, event cadence, community health metrics
- **Content Strategy.** editorial calendar design, format mix (tutorials, videos, talks, changelogs), SEO fundamentals for developers
- **Developer Events.** CFP strategy, talk submission, workshop design, conference presence
- **Metrics & Reporting.** DevRel KPIs, measurement frameworks, executive-facing dashboards
- **Developer Experience (DX).** API design reviews, docs audits, SDK onboarding flows, error message quality
- **Partnerships.** tech partner programs, co-marketing, joint launch coordination
Every playbook can be exported as Markdown and forked for team customization.
## AI-Assisted Tools
Four tools that generate structured, ready-to-use outputs from a short brief:
**Launch Checklist.** Input a product name, audience, and key dates. Get a prioritized pre/during/post-launch checklist with owner assignments and timing.
**Demo Run-of-Show.** Describe the demo scenario and get a minute-by-minute script with speaker notes, timing, fallback transitions, and audience interaction prompts.
**Content Repurposing.** Paste a piece of content and get a repurposing plan: which formats to adapt it into (thread, tutorial, short video, talk abstract), with draft copy for each.
**Demo-to-Distribution.** Turn a recorded demo into a full distribution strategy: YouTube description, Twitter thread, LinkedIn post, documentation hook, and newsletter blurb, all drafted from a single input.
## Resource Center
54 hand-curated resources across 6 categories: Developer Events, Communities, Learning, Tools & Platforms, Strategy & Metrics, and Reference. Every resource is annotated with a why-it-matters note. Community members can suggest additions, which go through a moderation queue before appearing.
## Community Layer
A space for DevRel practitioners to share tactics, link to playbooks, and surface what's working. All submissions are moderated to keep quality high.
## Who It's For
Developer advocates, DevRel leads, and developer marketers who want structured guidance and fast content production, without reinventing the wheel on every launch.
---
# Elden Ring Gallery
The Elden Ring Gallery is a web project that showcases artwork inspired by the highly anticipated game, Elden Ring. The gallery features a collection of stunning illustrations created by the talented artist "drakkar2guerre" from Instagram. The goal of this project was to create an immersive and visually appealing gallery experience for fans of Elden Ring.
**GITHUB:** PRIVATE REPO.
**DEMO:** 
## Search Reads the Plot, Not Only the Title
A keyword search finds the words you typed. Jeff finds the book you meant. Ask for "creepy floor plans and a murder mystery" and it returns *Strange Buildings* at 89%, *Strange Houses* at 82% and *Crooked House* at 80%, though no title contains those words.
The search runs on [Jev](https://typesafe.ai), a model from TypeSafe that answers typed questions with probabilities instead of text. It works in two rounds. A first round splits the shelf into chunks of 150 and asks Jev which item in each chunk fits best. A second round asks one yes or no question per finalist, so several books that all fit can all score high.
Every line Jev reads carries the book's plot, genres, characters and setting, pulled from each book's Goodreads page and from Open Library. A title cannot answer "books with dead bodies". A plot can.
## Anyone Can Make Their Own Shelf
Press **Make yours**, drop in a Goodreads export or a simple CSV template, and the browser builds your shelf:
- It finds covers on Open Library and samples a color from each one.
- Jev sorts every book into one of 21 categories.
- You pick one of 12 manga reader avatars.
- You choose what to publish: reviews, the to-read list, or single books stay private if you want.
Publishing gives you a link, a QR code, share buttons and a link preview card drawn from your own covers. Upload a newer export later and the link stays the same. A shelf you never publish stays in your browser.
## The Small Things That Make It Worth Sharing
- **Pick my next read** ranks your to-read list against the books you rated five stars.
- **Share this answer** turns any question into a link that runs it.
- **More like this** finds a book's nearest neighbours on the same shelf.
- **Reading Wrapped** draws your year in books as one image to save or post.
It works on a phone too, with bottom-sheet dialogs and no sideways scroll at 360px wide.
## How It Stays Cheap and Current
The site is static and runs on the Netlify free tier. A few Netlify Functions hold the Jev key, store published shelves, and log anonymous usage for an admin page. A search over my shelf costs $0.0067, and a daily budget switches search to keyword matching if a post ever goes viral.
My shelf updates from two sources. Books come from a Goodreads export, and articles come from the reading lists in my notes. One status command shows what changed for free. One build fetches covers, plots and categories only for items it has never seen, so adding a book costs a fraction of a cent.
[Ask my shelf a question](https://read.atharvashah.com), or make your own.
---
# Text Rhythm. The Cadence Visualizer
**Text Rhythm** is a zero-friction, in-browser prose analyzer that shows you the *shape* of your writing. Paste any text and instantly see how your sentences flow, their lengths, rhythms, and readability, without signing up, uploading anything, or sending a single character to a server.
## What It Does
Most writers know when something "reads slowly" or "feels choppy" but can't pinpoint why. Text Rhythm makes the invisible visible: sentence length becomes a bar chart, cadence becomes a waveform, and density becomes a heatmap.
## Five Views
**Sentences.** a live editable view. Write or paste directly. Sentences are highlighted as you move your cursor so you always know which one you're in.
**Paragraphs.** a bird's-eye view of your document by paragraph, showing how structure shifts across sections.
**Shortest.** surfaces your snappiest lines. Useful for finding punchy anchors to use as pull quotes or transitions.
**Longest.** flags runaway sentences that may be dragging reading pace.
**Heatmap.** every sentence is colored by length: green for short, neutral for medium, amber for long, red for anything that runs too far. Scan the whole document in one glance.
## The Rhythm Graph
The signature feature: a silhouette graph where each bar represents one sentence and its height is the word count. Monotone bars signal repetitive rhythm. A varied silhouette signals dynamic writing. Gaps between clusters show where your breathing points land.
## Stats Bar
Every keystroke updates the stats bar at the top: total words, sentence count, paragraph count, average sentence length, estimated reading time, and Flesch-Kincaid readability score.
## Focus Mode
Typewriter-style focus mode dims everything except the sentence under your cursor. Useful for reviewing one thought at a time without distractions.
## Three Themes
Dark, Paper, and Forest, because the right visual context matters for writing.
## Privacy
Your text never leaves your browser. Nothing is sent to any server. The editor persists your draft in `localStorage` so it's there when you come back. Clearing the editor is undoable (10-level undo stack, plus `Ctrl+Z` after Clear restores it).
## Who It's For
Writers, technical communicators, content marketers, and editors who care about sentence-level rhythm rather than word count. Especially useful when editing long-form blog posts, essays, or documentation where monotony creeps in.
---
# VocabCLI
VocabularyCLI is a lightweight Command Line Interface (CLI) that provides users with quick and convenient access to word definitions, examples, synonyms, and antonyms. It is a powerful tool for linguaphiles and language enthusiasts, offering a range of utility-based commands and robust knowledge base capabilities. With features like flashcards, tagging, word management, graph reporting, and bulk import/export of word lists, VocabularyCLI is a definitive software for expanding vocabulary and improving language skills.
The CLI boasts a simple and intuitive interface, making it easy to use for anyone. It offers advanced text classification and processing through the use of Natural Language Processing (NLP) and Machine Learning algorithms, which will be discussed in detail in the "Scope and Features" section. Additionally, the CLI provides an enjoyable and interactive user experience with eye-catching panels, tables, animated symbols, emojis, interactive menus, spinners, colored fonts, and other rich features.
**GITHUB:** https://github.com/HighnessAtharva/VocabCLI
**DOCS:** https://vocabcli.github.io/
## Project Tech Stack
- Primary Development Language: Python 3.10
- Database Management System: SQLite3
- API: DictionaryAPI (https://api.dictionaryapi.dev/api/v2/entries/en/hello)
- Deployment: PyPi (Python Package Index)
- Libraries: Rich, Typer, Matplotlib, Seaborn, NLTK, SpaCy, Pandas, Numpy, Requests, Beautiful Soup, PyDictionary
- Testing: PyTest (Unit Testing), PyTest-Cov (Code Coverage), PyTest-Benchmark (Benchmarking)
- Documentation: Sphinx, ReadTheDocs
## Learnings
During the development of VocabularyCLI, the project team gained valuable insights and learnings. Some of the key takeaways include:
- Understanding and implementing NLP techniques for text classification and processing.
- Utilizing various Python libraries for data visualization, data manipulation, web scraping, and API integration.
- Effective database management using SQLite3 for storing and retrieving word-related data.
- Deployment of the CLI package on PyPi for easy installation and distribution.
- Writing comprehensive documentation using Sphinx and ReadTheDocs to guide users in utilizing the CLI to its full potential.
- Testing the CLI using PyTest for unit testing, code coverage analysis, and benchmarking.
By applying these learnings, the project team was able to create a robust and feature-rich VocabularyCLI that meets the needs of language enthusiasts and provides a seamless user experience.
---
# What You Doing Right Now
WYDRN is a Social Networking website where you can track your movies, TV shows, games, books, and music. The aim of this project is to provide a centralized platform to track all your favorite media, eliminating the need to use multiple platforms like IMDB, Goodreads, Steam, RateYourMusic, and TVTime. It's one platform to rule them all!
**GITHUB:** https://github.com/HighnessAtharva/WYDRN
## Project Tech Stack
- FullStack Development with PHP and Bootstrap 5
- Login, Signup, Add Media, Import and Export, Diary, Browse Media, Edit Profile, Verify Account, Upload Image, Social Follow/Unfollow, and Feed functionality
- Database Schema Design and management of project file hierarchy
- API service integration and making API requests for all 5 media types (Movies, TV, Books, Music, and Videogames)
- Debugging, Pagination, AJAX, URL Rerouting, Version Management, SQL Queries, System Design, Caching, Security, Templating, Exception Handling
- Enforcing best practices and providing template/boilerplate code for ML Models
- Integration of Python with PHP for designing a Machine Learning model to recommend media types to users based on their added media items
- Scrum Master responsibilities including reporting project developments and ensuring successful builds for every iteration
## Learnings
Throughout the development of this project, I have gained valuable experience and knowledge in various areas including:
- FullStack Development, covering front-end and back-end implementation
- Database schema design and management
- Integration of API services for different media types
- Debugging and troubleshooting
- Implementation of security measures against XSS, CSRF, and SQL Injection
- Designing and integrating a Machine Learning model
- Scrum Master responsibilities and project management skills
With these learnings, I have successfully created a centralized platform for tracking and showcasing favorite media items across different categories.
---
# Sitemap
> Every page on atharvashah.com in one list, covering the main pages, the service pages, and the project pages, so nothing is more than one click away.
---
Canonical: https://atharvashah.com/sitemap
Last updated: 2026-10-26
---
# The Claude Code Slash Command Cheat Sheet
> Every Claude Code slash command and when to run it.
Every slash command, grouped by job and searchable, with a note on when to run each one and one tap to copy.
- Category: Claude Code
- Reading time: 3 minutes
- Items: 26
The full resource is free on the page after you enter an email address.
---
Canonical: https://atharvashah.com/resources/claude-code-slash-commands
Last updated: 2026-10-26
---
# How the 2026 LinkedIn Feed Ranks Your Posts
> What the 2026 LinkedIn feed reads and rewards, with a source for every claim.
What the 360Brew ranker reads, which formats win, and which advice is folklore. Every claim carries a source and a confidence note.
- Category: LinkedIn
- Reading time: 4 minutes
- Items: 17
The full resource is free on the page after you enter an email address.
---
Canonical: https://atharvashah.com/resources/linkedin-feed-2026
Last updated: 2026-10-26
---
# 30 Claude Code Tips From the Inside
> What decides cost, safety and speed in Claude Code, in 30 short tips.
How Claude Code loads context, spends tokens and decides what runs without you. Checked against the docs.
- Category: Claude Code
- Reading time: 16 minutes
- Items: 30
The full resource is free on the page after you enter an email address.
---
Canonical: https://atharvashah.com/resources/claude-code-30-tips
Last updated: 2026-10-26
---
# Privacy Policy
> What atharvashah.com collects, which services see it, how long it is kept, and the one control that switches analytics off for your browser.
---
Canonical: https://atharvashah.com/privacy
Last updated: 2026-10-26
---
# Terms of Use
> The rules for reading, quoting, and building on this site, plus what a listed price means before a written scope exists. Plain language, no filler.
---
Canonical: https://atharvashah.com/terms
Last updated: 2026-10-26
