Job searching is usually a second job you do for free. You comb through listings, open “apply” links, and then repeat the whole ritual until your brain turns into a spreadsheet. Autopilot-Jobhunt (A-J) aims to replace the drudge work with automation, using AI to scan the web for job openings, evaluate which ones match, and even help draft materials tailored to specific roles.

At the center of the project is Tarun Gupta, the software developer behind A-J. The bot is configured with a profile of the user, plus desired jobs and what the user “absolutely won’t accept” in an opening. Users can then let it run while they sleep: A-J scans the web, takes stock of positions that look like a good match, and sends a Telegram message listing all matching openings. Those openings are scored against the user’s resume and ranked according to the AI’s assessment. It can also format a resume and cover letter tailored to the position, but with a key constraint: users must review and send the application themselves, since the bot will not apply automatically.

That separation matters because a lot of the anxiety around AI in hiring is really about agency. A-J is built to assist, not to “take over.” The tool’s design suggests it is trying to land in the middle ground between speed and control: you move faster, but you still decide what actually gets submitted. Gupta also built privacy into the story, claiming the tool is designed to be private and including a full privacy readme in the project’s GitHub documentation. Under the hood, there are still real data flows worth noting for anyone thinking about operational risk: resumes are routed to the LLMs that OpenRouter is configured to use.

A-J’s tooling stack is also telling, because it reflects how these projects get assembled in 2025: not as a single monolithic model, but as a flexible chain of services and model providers. It uses TinyFish’s AI web agent to crawl for jobs. For the model layer, it relies on OpenRouter to run one of several default free AI models. The project starts with Llama and falls back to free versions of Nvidia’s Nemotron, Google’s Gemma 4, and Alibaba’s Qwen3 when the first option is unavailable or quotas run out. If someone has the tokens and wants an alternative path, Claude Code and the Anthropic API can be used in place of OpenRouter. Gupta said those who want to avoid sending resume data through OpenRouter can use Claude Code instead, provided they have an Anthropic subscription that supports it.

Now zoom out to why this is more than a clever side project. The source connects A-J to broader hiring dynamics, including last year’s reporting that researchers found some AI hiring bots, often the first line a company uses to separate “wheat from chaff,” favored applications generated by the same AI model used for screening. In plain English: if the evaluation system is biased toward a particular generation style, then the “human touch” may not matter as much as people assume. A-J doesn’t claim to solve hiring discrimination. But it does operationalize the workflow of generating tailored application materials fast, which could amplify that style-matching effect if companies rely on AI screening tools.

There’s also a market signal baked into the timing. The tool is configured by default for software developers, and the source ties that to Hiring Lab data published on Wednesday: the number of job openings for software developers has risen by 15 percent since Anthropic released Claude Code in February 2025, while openings for all other jobs have fallen by seven percent over the same timeframe. Even if you take that trend carefully, it still frames an important second-order effect for executives and boards: AI developer tooling ecosystems are changing not only how companies hire, but what kinds of roles applicants chase and how they find them. If job search behavior shifts, application quality and volume can shift too, which changes what screening systems see at scale.

Finally, there is the compliance and governance angle, because A-J explicitly does not auto-apply. That choice reduces the chance of “bot-submitted spam” or accidental submissions that can trigger reputational damage and platform enforcement. It also shows up in the project’s safer-by-design details: the config file where users link to their locally stored Markdown-formatted resume and set other options is gitignored so it won’t be committed by accident. Still, the resumes do get sent to the LLMs used via OpenRouter. That means privacy expectations will depend on model routing and provider selection, which is exactly the kind of thing that becomes a governance headache when teams scale AI usage beyond hobby projects.

For peers managing hiring funnels, A-J is a preview of an applicant-side capability leap. If more candidates can generate tailored resumes and cover letters quickly, it can increase the throughput of applications and compress the time window between “apply” and “screen.” That can push recruiters toward more automated triage and can increase pressure on any AI screening process to be accurate, fair, and explainable under real volume. For founders and investors, it’s a reminder that job searching is becoming another workflow subject to automation. For hiring leaders, it is a nudge that candidate tooling and company tooling are evolving in tandem, and the gap between them determines who gets advantaged and who gets missed.