Factory's Droid 1.0 lets an AI coding helper batch its steps into one script, which the company says cuts tokens, cost and wait time. The tests are unverified.
Factory has released Droid 1.0, a major update to the software that lets AI models act as coding helpers. Its headline feature, Script Mode, lets the AI do many small jobs in one go instead of one at a time. In one Factory test, a task that took 36 requests and 3.1 million tokens took two requests and 61,000 tokens.
Tokens are the small chunks of text an AI reads and writes. Fewer tokens means less work for the model. Until now, an AI agent used tools one step at a time, sending a request back and forth to the model for each step. Script Mode replaces that loop. The model writes a short JavaScript program, and that program runs many tool calls in a single request.
The script runs in a sandbox, a sealed space with no access to local files, networks or system settings. Big files and error logs stay inside the script. The model only reviews the final result, which keeps its working memory uncluttered.
In Factory's example, the job involved about 312 tool calls to move and verify 151 issues in Linear. Normal tool calling needed 36 requests. Script Mode needed two. Factory says Script Mode cuts overall session costs by about 15% on Claude Opus 5.5.
Factory also tested Droid on a benchmark called Terminal Bench 4, using the same models and tasks as each model maker's own agent software. These results have not been verified.
Factory shrank Droid's base instructions by 37%, from 20,887 tokens to 13,055. It reports a 98% prompt cache rate. Screens load about 3.8 times faster on startup, and the first prompt runs 3 times faster.
A new safety feature, Bounded Autonomy, spots when an agent is stuck repeating the same mistake and steps in. A new background design also lets users share active sessions across tools.
The material does not say what Droid 1.0 costs or how to get it. The performance claims come from Factory's own testing, and no outside group has confirmed them. Businesses that pay for AI coding tools may want to run their own trial before counting on the savings.
Factory says Droid 1.0 does the same coding work with fewer steps and tokens, but the figures are its own, so test it on your own work before trusting the savings.
EA has re-released 25 classic PC games at a discount until November 19, and Bethesda has finally given The Elder Scrolls VI a subtitle: Sentinel.
Read more: EA brings 25 classic PC games back, and Elder Scrolls VI gets a nameOpenAI released a ChatGPT for analysts and bankers that shows where each number came from, a feature that could matter well beyond finance.
Read more: OpenAI launches a ChatGPT built for Wall Street analystsA Kuala Lumpur firm now sells a SGD 1,500 check of what ChatGPT, Gemini, Perplexity and Google say about a named executive, after its founder's own test found mix-ups.
Read more: New Audit Checks What AI Chatbots Say About Business Leaders
