DEV Community: MyToolster

I Built a Life Simulator That Deals You a Random Country at Birth — With Real UN Data

MyToolster — Wed, 21 Oct 2026 05:10:35 +0000

Every life simulator starts the same way: you design your character. Pick a country, pick a family, optimize the stats. I wanted the opposite. Real life doesn't let you pick — you are dealt a country, a family, a body, at random, weighted by how many babies are actually born in each place. So I built a game around the deal itself: 1in8b.dev — "1 in 8 billion". You press one button, and you are born somewhere on Earth, with probabilities proportional to real birth counts. Then you live that life, year by year, until the local statistics say you don't.

The data layer

The whole game runs on public statistics for 172 countries and territories:

UN World Population Prospects for life tables and birth counts
WHO / World Bank indicators for life expectancy, under-5 mortality, income (GDP per capita, PPP), urbanization, fertility
Cause-of-death distributions per country — what people actually die of in Lagos is not what they die of in Osaka

That last one matters more than I expected. In the dataset, life expectancy runs from 54.5 years (Nigeria) to 85.5 (Hong Kong) — a 31-year gap determined almost entirely by the flag next to your birthplace. A simulation that ignores that is just a dice game. A simulation that uses it becomes something else: every run is a small lesson in how much of a "life outcome" is dealt before any choice is made. Mechanically, each in-game year rolls against the country's survival curve, its cause-of-death profile, and milestone events (schooling, work, migration, family) whose probabilities shift with the country's stats. Rich-country lives and poor-country lives don't just end at different ages — they have different shapes.

The build

It's a static site — vanilla JS, no framework, no backend, no accounts. All data ships as JSON and the simulation runs entirely in the browser. Two details from the build that might be useful to others:

1. Nine languages as static routes, not runtime switching. Each language is its own pre-built page (/ja/, /ko/, /fr/ …) with its own title, description, and canonical URL. The app detects the language from the path. This cost some duplication and bought a lot: every language version is independently indexable, and the first paint is already in the right language.

2. Country names are a real i18n problem. A UI string table is the easy part. But the game also needs the names of all 172 countries in every language — plus region names. We had Chinese, English, and Korean in the data; the other six languages were showing Chinese country names to French and Thai players. The fix was CLDR data (via Intl.DisplayNames) for the missing names, wired through one countryName() helper — and auditing ~250 hardcoded bilingual strings that had quietly bypassed the string table entirely. If your app "supports 9 languages," the long tail of strings outside the table is where it actually doesn't.

Try it

It's free, no sign-up, runs in the browser: 1in8b.dev

Play one life as yourself-by-birthplace, then re-roll and see who else you could have been. The second run is usually the one that sticks with you.

Burning Bridges AI: How One Photo Puts You in Drake's Party Scene

MyToolster — Mon, 12 Oct 2026 08:53:06 +0000

If you've been on TikTok or Reels in the last couple of weeks, you've probably seen it: someone raises a silver cup in a chandelier-lit room, scans the crowd with a hand over their eyes, and the whole party reacts around them. It's the Burning Bridges AI trend, and I finally sat down to figure out how it actually works — and how to make one yourself.

Spoiler: you don't need a video crew, a chandelier, or Drake's phone number. You need one photo.

What is the trend?

Back in May 2026, Drake released the "Burning Bridges" music video. About two and a half minutes in, there's a party scene: a dark, wood-paneled room lit by crystal chandeliers, Drake walking through a packed crowd with a silver cup, playing to a handheld camera.

In late September, people started recreating that scene with AI — replacing Drake with themselves. The clips spread under the name "Burning Bridges AI" on TikTok, Reels, and Shorts. The signature moments everyone copies:

The cup raise — leaning into the lens, finger-gunning the room
Working the crowd — turning, pointing, the room cheering back
"Where she at?" — a flat hand over the eyes, scanning left and right on the lyric. This is the move.

What makes it work is how specific the scene is. It's not generic party dancing — it's a particular room, particular timing, particular camera. That specificity is also what makes it hard to fake with a text prompt.

How it works (the unsexy version)

I tried the DIY route first: grab a reference clip, write a prompt describing the scene, feed it to a video model, hope for the best. The result? A guy in a vaguely fancy room doing vaguely dance-like things. The chandelier was there. The timing was not. The hand-over-eyes landed three seconds after the lyric. Not viral.

The tools that actually nail this trend do something smarter than text prompting:

A locked scene template. The room, the camera path, the crowd reactions, the second-by-second move list are pre-built from the real clip. You can't mess them up because you can't change them.
Character replacement. Your photo defines the lead: face, hair, build, outfit. The AI recasts you into the fixed performance instead of trying to invent choreography from a sentence.
Music laid back on. The finished video ships with the original track on the beat, ready to post.

That's the whole trick. The creativity is in who's in the scene, not in describing the scene.

How to make your own

Here's the actual workflow, using burningbridges.video as the example since it's the most one-photo-and-done option I found:

1. Upload one photo. One clear, front-facing photo with good light. Face sharp, upper body visible, one person per photo. This is 90% of your result quality — more on that below.

2. Pick your length and quality. The full scene runs about 25–28 seconds. Most tools offer a shorter cut too (around 15 seconds) for fewer credits. Resolution options are usually 480p / 720p / 1080p — higher res costs more, but for a phone-screen trend, 720p is the sweet spot.

3. Generate and wait. It takes a few minutes. The photo usually goes through a quick review first (face visibility check), then the render runs.

4. Preview, download, post. Check the face after the hand-over-eyes moment — that's where identity tends to wobble. Download the MP4 and add it to TikTok/Reels with the AI-generated-content label turned on (all three platforms ask for it now).

No prompt writing. No reference video hunting. No editing app. The whole thing is genuinely a few taps.

Tips I wish I'd known first

Your photo matters more than your settings. I ran the same scene with a dim bar selfie vs. a daylight photo. Night and day. Even, frontal lighting; face fully visible; no sunglasses, no heavy filters, no group shots. The AI can only work with what you give it.

720p is enough. I compared 720p and 1080p outputs side by side on a phone. For a vertical trend clip, the difference is marginal and the credit cost isn't. Save the 1080p for something you'll watch on a big screen.

Check the lyric sync moment. The "where she at?" beat is the make-or-break frame. If the hand lands wrong there, regenerate — most tools refund failed renders automatically, so a bad take costs you time, not money.

Label it as AI. Not just because platforms require it — the trend is about the AI swap. People engage with "that's me in the video??" energy. Own it.

Don't upload other people without asking. This should go without saying, but: only your own face, or someone who said yes. Most tools will ban accounts for non-consensual uploads, and rightly so.

Why this trend is interesting (beyond the memes)

I've watched a lot of AI video trends come and go, and most of them are prompt-lottery: type a paragraph, get a surprise, post the one good take out of twelve. The Burning Bridges format is different because the scene is fixed and the person is the variable. That constraint is what makes it reproducible — anyone's version is recognizably the same trend, which is exactly why it spreads.

It's also a preview of where consumer AI video is going: less "describe anything," more "pick a scene, insert yourself." Templates beat prompts for shareability, every time.

If you want to try it, burningbridges.video is the fastest path I found — one photo, a few minutes, done. I'm curious to see what people do with the format once the novelty wears off. My bet: friend-group versions (the tools that let you add background people) are the next wave.

Have you tried any of these AI video trends? What worked, what flopped? Drop a comment — I'm collecting notes for a follow-up on which trends actually hold up after the first week.

SEO Backlinks: Article Platforms Where You Can Publish Yourself

MyToolster — Wed, 07 Oct 2026 12:05:11 +0000

Directories are the slow lane for backlinks. Article platforms are the fast lane: you publish, you get the link, and good content gets indexed within days. Here are the platforms I've actually published on — with indexing speed ranking, how to boost indexing odds, and FAQ.

The lineup

Medium (https://medium.com/@mytoolster) — high domain authority. Profile bio with links, public Lists, engage under indexed articles.
Dev.to (https://dev.to/fastoolster) — developer-focused, great for tutorials. Canonical URLs supported.
Hashnode (https://mytoolster.hashnode.dev) — dev blogging with your own subdomain. Warning: AutoMod aggressively flags promotional "tool list" posts on free subdomains — write tutorials, keep same-domain links to 1-2 per post.
Substack (https://mytoolster.substack.com) — newsletter + blog in one.
Blogger (https://mytoolstertools.blogspot.com) — Google's own platform, indexes fast.
Hacker News / Reddit — not article platforms, but links in relevant discussions get crawled almost immediately. Engage genuinely; don't spam.
Qiita / Note / Zenn — Japanese tech platforms with strong SEO.
Your own blog — the hub everything points back to. Mine: https://mytoolster.github.io/

Google indexing speed ranking

Indexes very fast: Hacker News, Reddit, Dev.to, Substack. Indexes very slowly: Medium. So put time-sensitive content on the fast four; use Medium for the long game.

How to increase indexing probability

1) Comment under homepage articles on the same platform. 2) Cross-link your articles across platforms. 3) Link to your platform profiles from your own website footer. 4) Try different article formats. 5) Keep quality high.

Q: What if my account gets banned? A: Keep backup accounts and warm them up gradually. Don't put all content on one account.
Q: What if my article gets 0 reads? A: Increase indexing probability (five rules above) and keep publishing more high-quality articles. Indexing compounds.

Full version: https://mytoolster.github.io/posts/seo-article-platforms.html

MyToolster: 30+ Free Online Tools That Run Entirely in Your Browser (Including Video-to-SRT)

MyToolster — Wed, 07 Oct 2026 07:38:27 +0000

Like most developers, my bookmarks bar is a graveyard of single-purpose web tools: one for JSON formatting, one for QR codes, one for subtitle generation — each with its own signup wall, upload limit, or "free tier" that quietly isn't.

I got tired of juggling them, so I started consolidating the ones I actually use into a single collection: MyToolster — 30+ free online tools that run entirely in your browser. No accounts, no uploads to a server, no watermarks.

What it covers

The toolbox is organized into seven categories:

Audio & Video — speech transcription, video-to-text, and the standout: Video to SRT
Text — case converter, diff checker, word/character counters, Markdown to HTML
Data — JSON/JSON5 formatters, JSON to TypeScript, JSON ↔ TOON converters
Web — URL encoder/decoder and QR code tools
Color — color picker, Hex/RGB converters, Pantone converters
QR Code — generator and reader
Image — AI image detector, AI background remover

The one worth a closer look: Video to SRT

If you publish videos anywhere — YouTube, course platforms, social — you know subtitles aren't optional anymore. Most subtitle tools either upload your footage to a server or cap you at a few free minutes.

The Video to SRT tool works differently: it runs OpenAI's Whisper models (Tiny 39M, Base 74M, Small 244M) directly in your browser via WebGPU/WebAssembly. Your video never leaves your machine — it's demuxed locally, transcribed locally, and timestamped to the millisecond.

The workflow is simple:

Drop in any video (MP4, MKV, WebM, and more — even 4K)
Pick a Whisper model and the spoken language (or let it auto-detect)
Hit Transcribe and watch segments appear in real time
Export a standard .srt file and drop it straight into Premiere Pro, CapCut, or DaVinci Resolve

After the model is cached once, it works fully offline. No accounts, no usage caps, no "pro plan" upsell.

Why browser-local matters

The pattern across the whole toolbox is the same: compute happens on your device. That means:

Privacy — your files never touch a server
Speed — no upload/download round trips
Reliability — it works offline once loaded, and there's nothing to rate-limit you

It's a small thing, but it's the reason I reach for these tools instead of the tab graveyard I had before.

If you maintain your own set of go-to utilities, take a look at MyToolster and see if it can replace a few of your bookmarks. And if you publish video, try the Video to SRT generator — it's the tool I wish I'd had years ago.

AI Image Generators Still Can't Spell — Here's How I Fix Text Without Regenerating

MyToolster — Mon, 05 Oct 2026 11:07:24 +0000

AI Image Generators Still Can't Spell — Here's How I Fix Text Without Regenerating

You know the drill. You generate the perfect hero image in Midjourney — great composition, great lighting — and the sign in the background says "COFEEE SHOP". Or your OG image has the conference date wrong. Or a product mockup has last quarter's price baked in.

Regenerating is a lottery: you might get the text right on attempt #12, but the composition you loved is gone. Opening Photoshop for a two-word fix is overkill. There had to be a narrower tool for this.

The actual fix: edit the text in place

Instead of regenerating the whole image or rebuilding it in an editor, you edit just the text pixels. The workflow with PhotoTextEdit:

Upload the image (JPG/PNG/WebP)
Mark the text you want to change
Type the replacement
Download the result — background, lighting, and typography preserved

The tool inpaints the old glyphs away, reconstructs the texture underneath, and renders your new text matched to the original font style, weight, and perspective. It looks like the text was always there.

Where this actually saves developers time

OG/social images — fix a typo or update a date without re-running your image pipeline
Docs screenshots — update a label or version number in a screenshot instead of re-taking and re-annotating it
Demo mockups — swap placeholder prices, names, or dates on AI-generated product shots
AI art cleanup — fix the garbled text every generator still produces (yes, still, in 2026)

Honest limitations

It's AI-assisted editing, not magic. Results are best when the original text is readable and the change is focused — a word or a short phrase. Very small text, heavily stylized lettering, or text wrapped around complex 3D surfaces may need a second attempt. And obviously: don't use it on IDs, official documents, or anything meant to deceive.

The broader point

Not every image problem needs a full editor or a full regeneration. The tools that win are the ones scoped to one painful job and honest about what they can't do. Fixing baked-in text is one of those jobs — and now it takes about 30 seconds.

I used PhotoTextEdit for the examples above — free trial, pay per edit, no subscription.

I Run Claude Code on $0/Month: A Practical Guide to Free LLM APIs

MyToolster — Mon, 05 Oct 2026 10:45:28 +0000

I Run Claude Code on $0/Month: A Practical Guide to Free LLM APIs

AI coding assistants are incredible — until the API bill arrives. If you've ever watched your Claude Code or Cursor usage tick upward and wondered whether there's a cheaper way, this guide is for you.

The good news: in 2026, the free tier landscape for LLM APIs is genuinely good. Not "free trial for 7 days" good — permanently free, no credit card required, production-usable models. The bad news: the information is scattered across dozens of provider docs, and free tiers change constantly.

Here's the practical setup I actually use.

The free providers worth knowing

Not all free tiers are equal. These are the ones with permanent free tiers, no credit card required:

Provider	Standout model	Why it's good
Google AI Studio	Gemini 2.5 Flash (1M context, multimodal)	Most capable free tier, 500 req/day
Groq	Llama 3.3 70B, gpt-oss-120b	Fastest inference, ultra-low latency
NVIDIA NIM	100+ open models (DeepSeek R1/V3, Llama)	Huge selection, one API key
OpenRouter	35+ free models	Single key, single endpoint for everything
Cerebras	Llama 3.3 70B	Extremely fast
Mistral	Mistral Small/Medium	~1B tokens/month free

The full comparison — rate limits, context windows, credit card requirements — is maintained at freellm.net, a directory of 489 free models across 30 providers. I use it as my reference because free tiers change often and static lists go stale fast.

Step 1: Get a free API key (2 minutes)

Pick one provider to start. My recommendation for coding: Google AI Studio (best model quality) or Groq (best speed).

Go to aistudio.google.com (or console.groq.com)
Sign in with Google — no credit card asked
Create an API key, copy it

That's it. You now have a working LLM API key that costs $0.

Step 2: Point Claude Code at it

Claude Code reads two environment variables to override its backend:

export ANTHROPIC_BASE_URL="https://generativelanguage.googleapis.com/v1beta/openai"
export ANTHROPIC_AUTH_TOKEN="your-google-ai-studio-key"

(For Groq: ANTHROPIC_BASE_URL="https://api.groq.com/openai/v1".)

Then just run claude as normal. It will route through the free provider instead of Anthropic's API.

For Cursor: Settings → Models → Add Model, paste the base URL and key.
For Codex CLI: set OPENAI_BASE_URL + OPENAI_API_KEY.

Ready-to-copy snippets for every provider/model combination are at freellm.net/config — pick your tool, pick your model, paste.

Step 3: Verify before you commit

Before wiring a key into your daily workflow, test it. The freellm.net playground lets you paste a key and chat with any model right in the browser — no install, and your key goes directly to the provider, never stored on their servers.

Honest caveats

Free tiers have limits, and you should know them upfront:

Rate limits are real. Google AI Studio gives ~500 requests/day on Flash. Fine for personal coding, not for a team.
No free Claude. Anthropic offers no permanent free tier — you're running Claude Code the tool against non-Anthropic models. Quality differs; Gemini 2.5 Flash and DeepSeek R1 are the closest in coding ability.
Tiers change. Providers adjust limits quarterly. Check the current numbers before depending on one — this is exactly why a maintained directory beats a bookmarked blog post.

What this costs me

$0/month. I do the bulk of my AI-assisted coding on Gemini 2.5 Flash via Google AI Studio's free tier, and fall back to Groq when I need speed. When I hit a rate limit (rare), I switch providers — which takes 30 seconds since it's just env vars.

If you're a student, indie hacker, or just cost-conscious, there's no reason to pay for API access while these tiers exist. The setup takes less time than reading this article did.

Reference: freellm.net — 489 free LLM APIs from 30 providers, with rate limits, context windows, and one-click configs for Claude Code, Cursor, and Codex.