Downstream

Archives
Subscribe
September 12, 2026

๐Ÿ“ฑ Why did AI push Shopify back to native mobile?

Coding agents eliminated the cost of dual-platform maintenance

September 12, 2026

Coding agents eliminated the cost of dual-platform maintenance


The Deep End

How AI Agents Forced Shopify Back to Native Mobile Development

Coding agents eliminated the cost of dual-platform maintenance. This shifted Shopify's core tech stack back to native Swift and Kotlin after six years on React Native. Using AI workflows, engineers rebuilt their flagship Shop app in just 12 weeks. This analysis explains how AI translation tools erode cross-platform framework benefits and why decoupled app architectures enable faster AI-assisted iteration loops.

How AI Agents Forced Shopify Back to Native Mobile Development

AI coding agents fundamentally changed cross-platform software economics. Shopify is abandoning React Native to rebuild its apps in native Swift and Kotlin. Duplicate implementation no longer requires double the engineering effort. AI agents now handle cross-platform translation, testing, and parity checks reliably.

Team velocity exploded during initial native rebuilds. Engineers migrated the flagship Shop app to fully native code in just 12 weeks. Teams decoupled app logic into CLI tools so agents iterate in milliseconds without simulators. Direct platform integration ultimately outweighs cross-platform code sharing when AI handles translation.

Key Takeaways:

  • AI agents cut dual-platform maintenance costs, making native Swift and Kotlin development fast and cheap.
  • Decoupling app logic into desktop CLIs allowed AI agents to run tests in milliseconds.
  • Audit cross-platform framework costs to see if AI tools make native development more efficient.

Read the full article


The Periphery

Why Your Smart TV Scans Local Networks for Unconnected Devices

Security researchers caught LG smart TVs scanning local networks for unconnected phones and smartwatches. Unwanted network probes help LG Ad Solutions track 363 million secondary devices across American homes. This analysis explains how smart TVs map private networks and shows how to restrict background device discovery.

Your smart TV might spy on your phone. Security researchers found LG TVs scanning entire local networks for unconnected devices. The scans mapped staff smartwatches, phones, and neighbor Wi-Fi signals automatically. LG Ad Solutions uses this data to reach 363 million secondary household devices.

Key Takeaways:

  • LG TVs crawl local networks to build targeting profiles across 363 million secondary devices.
  • Wireshark traffic analysis revealed TVs gathering nearby Wi-Fi names, IP addresses, and signal strengths.
  • Isolate smart TVs on a segregated guest Wi-Fi network to block local device tracking.

Why Outsourcing Cognition to AI Destroys Value in Writing and Code

Outsourcing writing, coding, and thinking to generative AI replaces human judgment with unmaintainable average garbage. Author Ploum outlines a strict zero-AI policy: refusing LLM content restores critical cognitive practice and prevents long-term technical debt. This analysis examines why rejecting machine-generated output preserves intellectual discipline, protects software codebases, and elevates true human expertise in an automated world.

Delegating writing and coding to AI algorithms degrades human cognition. Author Ploum now rejects all machine-generated text, images, and software pull requests. Writing forces people to order contradictory thoughts into clear ideas. Skipping that mental labor wastes the reader's time and weakens the writer's mind.

Key Takeaways:

  • Delegating writing to generative algorithms erodes critical thinking by bypassing necessary cognitive ordering.
  • AI-generated code introduces hidden structural decay โ€” standardizing unmaintainable logic across software projects.
  • Require students and teams to submit raw prompts rather than AI-polished output.

How NASA Satellite Algorithms Reveal Invisible Ancient Rock Art

NASA originally built decorrelation stretch algorithms to map Martian terrain and volcanic lava flows. Today, archaeologists use that same color-expansion math to expose 200 invisible paintings across Angkor Wat. Matrix algebra expands color ranges instead of just boosting contrast. This analysis explains how space algorithms recover lost historical data across global sites.

NASA designed image algorithms to map volcanic lava flows from space satellites. A retired mathematician adapted the math into a fifty-dollar software tool called Dstretch. Archaeologists now use it to uncover 200 invisible temple paintings at Angkor Wat. Cross-domain technology transfers often solve stubborn problems in unrelated industries.

Key Takeaways:

  • Decorrelation stretch math expands faded color matrices to expose previously invisible archaeological art.
  • Simple contrast enhancement fails on ancient art because pigment wavelengths overlap with rock.
  • Audit existing technical algorithms in adjacent industries to solve your niche data challenges.

How Generative AI Tools Are Reshaping Newsroom Operations and Workflows

Generative AI tools accelerate routine background aggregation, freeing editors for original reporting. Journalist Nick Lichtenberg processes 50% of article drafts using AI tools. This workflow sparks intense media debate over editorial integrity. Learn how newsroom sandbox models drive successful technology adoption. Discover practical methods to maintain strict oversight and scale content production fast.

AI adoption in journalism raises urgent questions about content quality and labor. Fortune editor Nick Lichtenberg uses generative AI to outline features and synthesize documents. AI tools handle half the heavy lifting in gathering raw material. This approach forces editors to focus on primary reporting and human verification.

Key Takeaways:

  • Autonomous sandbox testing yields better AI adoption rates by encouraging organic editor experimentation.
  • Automated research tools process documents quickly, forcing journalists toward original field reporting.
  • Audit your editorial workflow today to isolate routine synthesis tasks for automation.

The Firehose

Autonomous AI Agents

  • How Agentic AI Loops Automate End-to-End Data Journalism Projects
  • Why Autonomous AI Coding Agents Generate High-Volume Unreadable Software Slop

Worth Exploring

  • How Local Black Newspapers Survive Sixty Years Without Losing Trust
  • Why Keeping Legacy Cables Is Actually a Rational Tech Strategy


View this email in your browser ยท Browse past issues

Don't miss what's next. Subscribe to Downstream:
Older โ†’ ๐Ÿ“Š What publisher adoption reveals about AI audio
Powered by Buttondown, the easiest way to start and grow your newsletter.