Sponsored
vercat PokerGod Token Reset 3 Things Today
$5/week ↗
Projects

I Called My AI Agent. It Knew My Threads Numbers.

@lennoxsaint

added

I called my AI agent and asked how my Threads were going. It read my real numbers back: +653 followers in 30 days. Then one prompt set it up to learn my voice and draft my week.

Lennox Saint calls his Dot and asks how his Threads are going; it reads his real numbers back on the call (+653 followers in 30 days; his own snapshots: 31,905 on Aug 31 to 32,558 on Sep 30, 2026). One prompt then sets it up as his Threadify social-media operator: it learns his voice from his notes and drafts his week for review, nothing posts without him. First-hand on-camera call with his own account data. Disclosed: video funnels the creator's own Threadify product (threadify.app plans link in description); creator notes the same one-prompt setup also runs in the ChatGPT app, Claude, OpenClaw and Hermes, framing this video's run as the OpenAI Dot. Real handle: @lennox_saint (validator-safe handle used).

View on YouTube

More Dots use cases in Projects

See all 294 →
OpenAI Dot stress-tested with a hiring task

OpenAI Dot stress-tested with a hiring task

I Gave an OpenAI Dot a Hiring Task, Then Tried to Trip It Up I gave an OpenAI dot 4 fake intern applications to organize. Then I changed the priority, added 2 more, and slipped in a duplicate. Could it keep it straight? Quick context: dots are OpenAI's always-on agents. Each has its own cloud computer and can work through ChatGPT Work or Codex. The test was all fictional. Rules: organize the evidence, keep gaps visible, leave decisions to me. And it got interesting fast. Applicant two claims Python, but the project was a browser interface with no Python sample. The dot kept both facts visible and marked the evidence unknown. It did not turn a claim into proof, and an absent sample just tells you what to ask for. Changed priority to project-first: it reordered the columns, changed nobody's eligibility. Added 2 more plus a duplicate: got 6 unique applicants, not 7. Asked for an interview draft: 5 project-specific questions, left as a draft. It did not pick the applicant. I did. Honest part: one staged session, fake data, no comparison to normal chat. It organized the material. The decisions stay mine. Full breakdown: https://openai.com/index/introducing-dots/
LearnOpenCV
Dot planned her LA trip live: refundable flight + hotels, then compiled user feedback

Dot planned her LA trip live: refundable flight + hotels, then compiled user feedback

A woman at her desk gives an unfiltered look at how she actually uses her Dot (framed as a remake of OpenAI's failed DevDay demo 'with better WiFi'): she voice-calls her Dot about an important LA meeting tomorrow around 3pm, asking for a policy-compliant refundable flight and East Hollywood hotel options. Dot reports a Southwest 11:30 departure landing 12:55, a refundable fare option around 3:26, a 30-45 minute LAX exit estimate, and hotels including Silver Lake Pool & Inn and Cara (she checks the browser history showing those hotels). She then asks Dot to identify top themes from her user feedback and propose solutions, and Dot confirms it will compile themes, fixes, and gaps into a concrete workflow.
@louis_enda