Chirath Soithong GitHub Résumé PDF

Side projectBuilt and run by me2026

A short news podcast on my phone, every weekday morning.

BeforeI scrolled several news sites every morning.AfterOne message arrives on its own: a short audio briefing, a recap and the links.

ResultRunningIt runs by itself every weekday, after the market opens.

Jump to the resultSee what arrivesTechnical deep dive

Close-up of hands opening the business section of a morning newspaper
Illustration: the morning business pages.Photo: Olu Eletu (Unsplash), CC0, via Wikimedia Commons

The 60-second version

What I built
A small automation that collects the day’s news, has an AI model pick what matters and write a short script, and has a natural-sounding voice read it. The audio, a recap, photo cards and the links land in my Telegram.
  1. A timer starts it
  2. Collect the news
  3. AI picks and writes
  4. A voice reads it
  5. Sent to Telegram
  6. I listen
Every step runs on its own. I choose what to read in full.
Who did what
The idea, the quality bar and the design were mine; AI helped me research each tool first. AI models write the summaries and read them aloud, inside rules I set. Every story links back to the original article.
Runs as
A timer, every weekday around 11 am New York time: after the opening bell, so there are articles to summarise. Nothing to press.

Part 1 of 4: The problem

I wanted the day’s news without opening several sites every morning.

Markets, tech and world news in one place. If a machine was going to summarise it for me, it had to be good enough to listen to every day.

Exhibit 1My morning news, before and now

Drag to compare

Illustrative picture. Before: four browser windows, a finance news site, a newspaper, a tech news site and a world news site, each full of headlines to read. Now: one message from my briefing bot in Telegram, with the day’s headline, a short audio briefing (3 minutes 40 seconds in this example), a recap, photo cards and all the day’s stories.

Illustrative The shape of a real run (Sep 16, 2026), shortened.

Part 2 of 4: What I built

One message each weekday: press play, or read the recap.

Around 11 am New York time, after the opening bell, it runs on its own and sends the package to my Telegram.

Exhibit 2What lands in my Telegram each weekday

Point at a number

Illustrative picture of one weekday message, the shape of a real run on Sep 16, 2026, shortened. Numbered parts: 1, the audio briefing, 3 minutes 40 seconds. 2, the recap: policy, markets and tech, one line each. 3, two photo cards, each with a headline, one line on why it matters and a link. 4, all the day’s stories, each with its source. The list under the picture explains each part.

Finance Briefingbot

Global briefing · Sep 16

Fed hikes, Dow drags in a mixed session

3:40

PolicyThe Fed raised rates a quarter point, its first hike in three years.

MarketsMost indexes slipped a little; the Dow fell harder, dragged by a few big names.

TechA public jab between two AI labs made the front pages.

10:59 AM

Photo cards

The Eccles Building, headquarters of the US Federal Reserve in Washington, DC, under a blue sky

Fed raises rates for the first time in 3 years · SCMPWhy: the Fed–White House tension, first hand.read article

Gloved hands holding a finished silicon wafer in a cleanroom

China runs a weather model on 100,000 home-grown AI chips · SCMPWhy: domestic chips at scale, despite export controls.read article

10:59 AM

Global · all stories

10:59 AM
  1. The audio. A short spoken briefing, read by a voice I chose after testing others.
  2. The recap. What moved the markets and why, in the same short shape every day.
  3. Photo cards. The top stories, each with one line on why it matters and a link.
  4. All stories. Every item names its source and links to the original article.
Illustrative The shape of a real run (Sep 16, 2026), text shortened. The card photos are open-licence stand-ins; the real cards carry the articles’ own photos. A blue dot marks a top story.
How one weekday run works, in five steps1 exhibit · under 1 min

Exhibit 2.1One weekday run, step by step

  1. A timer starts itWeekdays, around 11 am New York time. Nothing to press.
  2. It collects the newsStories from several news sources: finance, technology and general news.
  3. An AI model picks and writesIt picks what matters and writes a short script.
  4. A voice reads itText-to-speech turns the script into audio.
  5. Telegram gets the packageThe bot posts the audio, the recap, the photo cards and the links.
  6. I decide what to read in fullEvery story links back to the original article.
Green marks where a person decides.

Part 3 of 4: My choices

I tested each part before it went into the daily message.

I listened to different voices, compared image styles and tried AI-made charts. Then I kept tweaking the format, run after run.

The bar I setWorth listening to every day.

Exhibit 3Three things I tested

Pick a test
  1. The voice

    I listened to different text-to-speech voices and kept one that takes notes on tone and emotion, so the briefing sounds conversational, not robotic.

  2. The pictures

    I compared image tools that can run inside an automation, from cinematic to cartoon-like, against the look I wanted for financial news.

  3. The charts

    AI-made charts looked convincing, but their numbers weren’t reliable.

Illustrative Sketches of each test, not real output. The tone words are examples.
My part, and what the AI does2 short lists · under 1 min

My part

  • The idea and the bar: worth listening to every day.
  • Designing the steps and wiring the tools together.
  • Choosing the voice after testing options, and tuning its tone.
  • Tweaking the format, run after run.

What the AI and the timer do

  • AI models write the summaries and read them aloud, inside rules I set.
  • AI helped me research each tool before I built with it.
  • A timer starts every run; I don’t press anything.
  • The journalism is the newsrooms’: every item names its source and links to it.

Part 4 of 4: What changed

Running

It runs by itself every weekday. I listen instead of scrolling.

Built for one listener: me.

SkillsWorkflow automation · Testing AI tools · Setting a quality bar

Exhibit 4What I stopped doing by hand

Scroll to cross out the old way
  1. Repetition

    Before: Open several sites, every morning.

    Now: The summaries and the links arrive in one place.

  2. Hand-starting

    Before: Remember to go and look.

    Now: Nothing to press. It runs itself on weekdays and tells me if a step fails.

  3. Reading

    Before: Read everything to find the few that matter.

    Now: Listen for a few minutes. Open a link only when I want more.

Built with

  • Python, collects the news
  • n8n, the timer and the steps
  • Claude, picks and writes
  • Inworld, the voice
  • Cloudflare R2, stores the audio
  • Telegram, delivers the message