AI Week Radar

issue #002 · May 22, 2026

ChatGPT wants your bank login, AI runs radio, junior devs run scared

OpenAI connects to your checking account; Andon Labs hands radio stations to LLMs

OpenAI connects to your checking account; Andon Labs hands radio stations to LLMs
AI Week Radar #2
Fri, May 22

OpenAI is now asking for your bank credentials, Andon Labs handed entire radio stations to language models, and Fortune notes the entry-level job market is already thinning. Meanwhile the tooling crowd keeps debating whether to write code at all.

— Jarek

Featured

OpenAI launches ChatGPT for personal finance, will let you connect bank accounts — the new feature aggregates linked accounts into a dashboard of "portfolio performance, spending, subscriptions, and upcoming payments." An LLM with read access to your checking account is a compliance, PII, and prompt-injection surface unlike anything OpenAI has shipped before. Watch how scopes, fraud detection, and regulator pushback play out — this is the template every assistant vendor will copy or avoid. — Ivan Mehta

Articles & Videos

We let AIs run radio stations — Andon Labs handed music selection, live mixing, and listener interaction to autonomous agents. The interesting parts aren't the playlists — they're the latency budgets, error recovery, and state management needed to keep a multi-modal agent on-air without dead silence. A useful field report for anyone shipping long-running agents in production. — lukaspetersson

INDUSTRY

AI is wiping out entry-level jobs — QA, junior dev, and annotation roles are the obvious automation targets, and Fortune argues the experience gap is widening fast. Founders rethinking headcount and engineers planning the next decade both have a stake here — the pipeline that produced today's seniors is being quietly dismantled. — Brajeshwar

DEEP DIVE

From Vibe Coding to Spec-Driven Development — a fitness app shipped in 4.5 hours by writing specs instead of code, then letting agents do the typing. The piece is a decent maturity check on spec-to-code workflows: where they hold up, and where the agent still needs a babysitter. — Mariya Mansurova

DEBATE

Going full AI engineer, not touching code anymore — one engineer's account of fully delegating implementation to LLMs — and the productivity, debugging, and maintainability trade-offs that come with it. Pair it with the spec-driven piece above for two takes on the same question: what's left for the human? — maxheyer

Releases

InsForge – Open-source Heroku for coding agents — an open-source deployment layer for coding agents, pitching itself as the Heroku abstraction for agent workflows. Early days (43 HN points, 6 comments) so kick the tires before betting production on it. — mrcoldbrew

Last but not least

How I Continually Improve My Claude Code — a practical write-up on feedback loops and prompt refinement for Claude-driven coding. Light on benchmarks and cross-model comparison, but the patterns travel. — Eivind Kjosbakken

See you next week — try not to outsource everything.

You're receiving this because you subscribed at aiweekradar.com.
Read on web · Manage preferences · Unsubscribe in one click

© AI Week Radar · 8cells Jarosław Kaczmarski · ul. Ketlinga 33a, 96-313 Jaktorów, Poland