SShortSingh.
Back to feed

A 16-year-old runs his company with AI agents and logs nine failures from 453 sessions

0
·4 views

A 16-year-old founder says his business is run mostly by Claude Code sessions that write code, handle supplier communications, publish content and monitor websites. He measured 453 sessions from October 7 to November 11, 2026, covering 17,050 tool calls and 38.3 hours spent inside tools. The article details nine failures, including foreground commands that froze his terminal for up to 68 minutes and a duplicate message sent to three suppliers after a faulty success check. Each failure produced a rule, such as running long tasks in the background and rereading real state before retrying a send. He also found that his agents loaded only 17 of 148 skills, so he rewrote descriptions and plans to delete unused ones. The full measurement script is publicly available on GitHub.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Orbit: A Free, Self-Hosted Personal Assistant That Replaces Ten Apps

A developer built Orbit, a free and open-source personal assistant designed to run on the user's own machine. It consolidates tasks, calendar, finances, notes, habits, documents, health data, and more into a single dashboard with unified search. An optional AI assistant is disabled by default and can run locally through Ollama or use the user's own API keys from providers such as OpenAI, Anthropic, or Gemini. Its agent workflows require user approval for every change and lack any tools for deleting, paying, emailing, or sharing data. The project is available on GitHub and can be launched with Docker Compose, using SQLite by default or PostgreSQL as an alternative.

0
ProgrammingDEV Community ·

EU Set to Ban Dark Patterns in Web Design, With Penalties Up to 6% of Turnover

The European Union is preparing regulation that would outlaw deceptive interface designs, known as dark patterns, that mislead or pressure users into commercial actions. The proposed rules, reportedly targeting entry into force on January 1, 2025, would cover digital services offered to EU consumers, including e-commerce, apps, and SaaS tools. Penalties could reach 6% of global annual turnover or €30 million, whichever is higher. Industry commentary suggests many violations can be fixed with incremental interface changes rather than full redesigns, though some companies may see short-term conversion dips during the transition.

0
ProgrammingDEV Community ·

Dev.to Users Propose Customizable Profile Pages Like GitHub READMEs

A DEV Community member observes that author profile pages on the platform look bare and make it hard for writers to stand out. They propose adding a feature similar to GitHub's username/README.md, which renders a special page on a user's profile. Such a page would let authors include graphics and richer presentation rather than relying only on a plain bio. The post suggests two possible implementations: an article titled with the author's username, which could be a draft and would be pulled onto the profile automatically, or letting authors select an existing article in their settings. The post seeks community feedback on the idea.

0
ProgrammingDEV Community ·

Freelance seller turns a cheap VPS into a shared inbox to stop missing customer messages

A solo Vietnamese seller lost a sale after overlooking a loyal customer's Zalo message during a busy pre-Tết period. The incident showed that inquiries were scattered across Zalo, Facebook and email with no single place to track unanswered messages. To fix this, the seller set up a low-cost VPS in Vietnam for data control and low latency, hosting a shared mailbox and a request-tracking sheet. Each incoming message is logged manually, with a deadline and status, and requests are grouped into urgent, normal and can-wait tiers with color coding. The approach relies on simple tools and deliberate logging habits to ensure no customer request is missed.