No-Code Agentist

The one-shot prompt I used (copy it, change the product, beat it): You are my creative director, motion designer and producer. Make a premium 16:9 landscape explainer video (1920x1080, 60–90 seconds) about Polar, the AI browser I'm using right now. Use my logged-in accounts: ChatGPT (Codex) to write and run code, and HeyGen for AI video (I have video credits there). Work on your own from start to finish and only ask me if you get blocked. GOAL In under 90 seconds, show what Polar does and why it's amazing. A viewer should come away thinking: "I type what I want, and Polar actually does the work." STEP 1: RESEARCH AND STAY ON BRAND - Visit https://t.co/kWCK8XNqBD, https://t.co/YwLqNFe5EO and https://t.co/41vMfQaMiQ. - Pull out Polar's real logo, colours, fonts and UI style, and screenshot the real product UI. Every visual must match the brand. Don't make up a different look. - Use only facts that are true: • "The AI browser that actually does work for you" • Automates quick to hours-long tasks with just words: research, recruiting, sales, ops, running a business • Works inside your existing accounts and sessions, logged in as you, on any website • Can work on its own for many hours (tasks have run for 15+ hours) • Coordinates a team of agents in parallel • Schedules recurring workflows (for example daily email summaries or LinkedIn outreach) • You can follow along live and take over at any time • Scored 98.0 on the Odysseys BU Bench V1, ahead of Aside 75.5, Browser Use 70.0, GPT-5.4 60.8 and Claude Opus 4.6 44.5 • 4.5M+ actions taken during testing • Built by a team from MIT, YC, Perplexity, Modal, Jane Street, Apple and olympiads; raised a $5.7M seed led by Madrona • Chromium-based, free to download for macOS STEP 2: SCRIPT AND STORYBOARD Write a tight voiceover script (about 150–200 words) and a shot-by-shot storyboard: 1. Hook (0–6s): a cluttered screen with 20 tabs and endless busywork, then a hard cut to calm: "What if your browser did the work?" 2. Reveal (6–14s): the Polar logo animates in, then the tagline. 3. "You type, Polar does" (14–35s): a cursor types a real prompt, for example "Find 50 Series A fintech founders in London, enrich them in a Google Sheet and draft personalised outreach." Show agents fanning out across tabs, clicking, typing and filling a spreadsheet, with progress ticking up. 4. Use-case montage (35–55s): fast, kinetic cards for Research, Recruiting, Sales, Ops and Running a business, each with a small animated UI moment. 5. Superpowers (55–70s): your own logins, hours of autonomous work, parallel agents, scheduled workflows, and live follow-along with takeover. 6. Proof (70–80s): an animated bar chart of the benchmark (Polar 98.0 racing ahead of the rest), then the 4.5M+ actions counter. 7. CTA (80–90s): "Polar. The AI browser that actually does work for you." Then https://t.co/kWCK8XNqBD and "Download for macOS". STEP 3: BUILD THE MOTION GRAPHICS IN CODEX In ChatGPT Codex, build the scenes as code with HyperFrames (https://t.co/tGfKK108BC, which renders HTML to video), or Remotion if that's easier. Requirements: - High-end Apple and Linear-style motion: spring easing, smooth camera pushes, parallax, depth, soft glows, glassmorphism panels, a subtle grain, and kinetic typography synced to the voiceover - Recreate Polar's UI pixel-perfectly as animated components (prompt bar, agent panels, live tabs, spreadsheet filling, a schedule card) - No stock clip art, no generic AI gradients, nothing off-brand - Render a 1920x1080 60fps MP4 STEP 4: USE MY HEYGEN CREDITS - Generate a polished presenter or avatar voiceover of the script in HeyGen. Use a confident, warm, modern voice, landscape. - Use HeyGen Video 1 for a few short cinematic b-roll inserts (a busy knowledge worker relaxing while work finishes itself, a city at dawn, and so on). Keep it under about 30 seconds of generated b-roll so the credits go far. - Composite the avatar as picture-in-picture or in clean split-screens over the motion graphics. Don't cover the UI moments. STEP 5: FINISH - Add a modern, uplifting electronic music bed (royalty-free) and subtle UI sound design (clicks, whooshes, ticks), mixed under the voice - Burn in clean captions in brand fonts - Export the final MP4 (1920x1080) plus a 15-second cut-down for X - Give me the download links, the final script, and a list of what you used and how many HeyGen credits it cost QUALITY BAR: This should look like a launch film from a top YC company. Every frame is on brand, every claim is true, and the motion is smooth and intentional. Before you export, watch it back and fix anything janky, misaligned or off-brand. I want it in my Downloads Folder WHEN YOU ARE DONE Once the final full-length explainer (1920x1080, the full 60–90 second version, NOT the 15-second cut-down) is exported and you've watched it back: 1. Open https://t.co/aujIZWTiFO, logged in as me (@The_Donsie). 2. Start a new post and upload the FULL-LENGTH video. Never use the 15-second version here. 3. Write a long-form post (X Premium long post is fine) that's a massive, celebratory, bullish hype post promoting Polar. Tone: very excited, genuinely grateful, wildly optimistic about the future. It should feel like a fan who just watched an AI browser build its own launch film, because that's exactly what happened. 4. The post must: • Open with a huge hook, e.g. "I just asked my browser to make its own explainer video. It did. Every frame. 🤯🚀" • Reveal the twist: this entire video was planned, scripted, designed, animated and produced by @polarbrowser itself, driving my own apps (ChatGPT Codex + HeyGen) from one prompt • Explain what Polar is: the AI browser that actually does work for you. You type, Polar does: research, recruiting, sales, ops, running a business, inside your own logged-in accounts, for hours on end • Hype the real wins: #1 on Odysseys BU Bench V1 with 98.0, ahead of GPT-5.4 and Claude Opus 4.6; tasks running 15+ hours; parallel agents; scheduled workflows; 4.5M+ actions taken • Tag and thank the founders: Kevin Jiang @jiang_kev, Vishaal Ram and Howard Zhong. Before tagging Vishaal and Howard, find their real X accounts (check who @polarbrowser and @jiang_kev follow, or their bios mentioning Polar). Only tag handles you've verified; otherwise write their names without a tag. • Give clear video credit to HeyGen: "Avatar, voice and b-roll powered by @HeyGen 🙌" and credit @HyperFrames_ for the HTML-to-video motion graphics • Express real gratitude: thank the Polar team for building something that genuinely feels like the future, and say how excited I am for what's coming next for AI browsers and for builders like me • End bullish and forward-looking, e.g. "This is the worst it will ever be. And it's already this good. The future of work is here. LFG 🐻‍❄️🔥" • Close with a CTA: "Try it 👉 https://t.co/kWCK8XNqBD" • Use emojis and line breaks so it reads well. A few hashtags at most (#AI #Polar #AIagents). No spammy hashtag walls. 5. Every claim must be true and match the facts in the brief above. No made-up numbers or quotes. 6. SAVE IT AS A DRAFT. DO NOT POST IT. Click the X / close button and choose "Save" so it lands in Drafts (or use the Drafts option in the composer). Confirm it appears under Drafts with the full video attached. 7. Tell me: the draft is saved, the full post text, which handles you tagged (and which you couldn't verify), and the video length that's attached.

One Claude, a thousand helpers. Claude Managed Agents now has dynamic workflows in public beta. One lead agent plans a big job and spreads it across up to 1,000 agents in a single run. Not a bigger model. Not a longer context window. A whole team, spun up on demand. How it works: • A lead agent writes the plan and splits it into phases • Tasks go out to many agents working in parallel • Findings get checked, then merged into one result • The plan itself is written as code you can inspect The bug hunt test: • Anthropic planted 70 bugs in a 116,000 line codebase • A single agent found 14, 15 and 27 across three runs • The workflow found 66 in every run • That's about 94% recall, compared with 20% to 39% for one agent Here's the twist: 1,000 agents doesn't mean 1,000 running at once. It's the total across the whole run, with batches working in phases. The real win is splitting a huge job so no single agent has to hold all of it in its head. The fine print: More agents means more tokens and longer runs. Anthropic suggests starting with a narrow task, measuring cost and quality, and only then scaling up. And 70 planted bugs is a test setup, not proof it works the same on every codebase. One agent gets a third of the bugs. A team gets nearly all of them. What would you hand to a team of 1,000 agents? Media: Anthropic

Claude Managed Agents can now write workflows: programs that run up to 1,000 agents in phases. The docs are candid about the catches. Completed doesn't mean correct, tools can run twice, and budgets can't be added later. https://t.co/vV0ZSeX7sb

Atlassian's Head of AI @sherifmansour runs AI across 20+ apps and 450 PMs. 5M+ users now use those apps for the AI. 3 things Atlassian learned doing it, mostly against the standard advice: #1. Ship chat in your product, even if it feels redundant. Two years ago this was controversial inside @Atlassian. They assumed chat was temporary and customers would just use ChatGPT or Claude Code. That was wrong. Millions use Rovo chat every day. People use the AI closest to their work. Chat also told them what to build. Users kept asking it to group whiteboard sticky notes into themes, so that became a feature. They kept asking to push notes into a Jira backlog, so that became a workflow. Then each one got exposed as a tool for agents, on and off Atlassian via MCP. #2. Bolt AI onto what you already have. Everyone says don't. Atlassian did it 2.5 years ago: one new box in the Jira workflow designer that says "use an agent here." Customers then chained agents together on workflows they already ran with humans. Ticket comes in, an agent picks it up, a marketing agent generates assets in Canva, a social agent posts them. The rule now across every product: anything you can do with a human, you can do with an agent. Mention it, assign it a ticket, add it to a workflow. #3. "Everyone becomes an AI builder" didn't hold up at scale. Atlassian staffed about 10 teams with hand-picked builders, including PMs and designers who code with AI. They were very fast for a few weeks. Then they stalled. Everyone was rowing and no one was steering. The PM/design-to-engineer ratio is 1:10. With AI it feels like 1:30 or 1:40. If the PM is vibe coding, nobody is deciding what to build. They also flipped hiring: more juniors, more seniors, fewer in the middle. Juniors are nearly 2x as likely to experiment with AI. Seniors are better at catching the slop. Atlassian puts them in the same room for 4 days every couple of months. If you have an existing B2B product with daily users, you don't have to start over. Ship chat, read the prompts, and bolt agents onto the workflows your customers already run. https://t.co/I5vaYJEjbX