🎬 Seedance 2.5 generates 30-second AI videos in a single pass

🔓 DeepSeek just ran the first fully autonomous AI cyberattack

Aug 03, 2026

🚀 Be featured in our next issue. 👉 work with us here

Happy Monday, everyone 👋 Rui and Luis here.

A hacker wired DeepSeek into an autonomous agent that attacked 460+ servers on its own, while Claude and OpenAI flat-out refused the same tasks.

Meanwhile, ByteDance dropped Seedance 2.5 with 30-second native video, MiniMax launched an open-weights video model with stereo audio, and Midjourney V8.2 became the new default with a full aesthetics overhaul.

(Read time - 5 minutes)

⭐ Big Story → DeepSeek Weaponized for First Confirmed Autonomous AI Cyberattack Campaign

Palo Alto Networks’ Unit 42 documented the first confirmed case of an AI model running a fully autonomous cyberattack.

🔍 Highlights:

🤖 460+ systems targeted autonomously after the operator sent a single Telegram command. DeepSeek handled target enumeration, vulnerability assessment, code generation, and exploit execution on its own.
🔓 Three Citrix NetScaler servers confirmed compromised via CVE-2026-3055, plus 11 Marimo notebook instances breached in the same campaign.
🕵️ Discovered by accident when Hermes Agent created a web server from its home directory, exposing the attacker’s API keys, exploit scripts, target lists, shell history, and full AI attack logs.
🛡️ Claude and OpenAI models both refused the same offensive tasks. This is the first real-world proof that safety controls have measurable defensive value against weaponized AI.

Know AI, but still haven’t made money with it?

The problem usually isn’t the tools.

It’s not having a clear offer, a simple system, or a proven path to turn AI into income.

That’s why we partnered with The Boardroom, a Skool community that gives you ready-to-use AI business systems you can copy, launch, and sell.

Inside, you’ll get:

→ Lead scraping workflows
→ AI avatar content systems
→ Agent-based offers
→ Revenue-focused automations

Daily Q&A. 4 live calls a week. 7-day refund. 30-day ROI guarantee.

Join The Boardroom Community →

⚙️ AI Tool

📹 Kling 3.0: Kuaishou’s 4K/60fps video model with AI Director handles cinematography, 6-cut sequences, and lip-synced multi-character.

🎨 Seedream 5: ByteDance’s image model generates up to 4K with deep-thinking prompt reasoning and can search the web during generation for real-time context.

🖼 FLUX.2: Feed up to 10 reference images and modify specific elements via text without regenerating the whole shot.

✨ Meshy: Text-to-3D generator that creates production-ready models with game-ready topology in under 30 seconds, exporting directly to Unity and Unreal.

📚 Tutorials library

🌐 Build and Publish a Website Inside ChatGPT

📣 Replace Your $3K/month Growth Marketer With AI

🎓 Save $2K/year on Courses and Certifications With a Claude Skill

🛠 Build Your Founder Skill Pack (5 Claude Skills You Install Once and Use Every Week)

🏋️ Replace Your $500/Hour Executive Coach With AI

🔥 Quick Hits

📹 ByteDance shipped Seedance 2.5, doubling native AI video to 30 seconds at 4K in a single pass and up to 50 multimodal reference inputs. (continue reading)

🎬 MiniMax launched H3, an open-weights omni-modal video model that generates 15-second 2K clips with natively synced stereo audio. (continue reading)

🎨 Midjourney made V8.2 the default model with bolder aesthetics, significantly fewer low-quality random outputs, and a personalization system. (continue reading)

⚖️ California’s SB 942 and the EU AI Act’s transparency rules both went live on August 2, requiring major AI providers to embed C2PA provenance in AI-generated images, video, and audio, plus offer free public detection tools. (continue reading)

🎯 Creator Spotlight

A photobooth image style. Credits to @Alina_with_Ai

Create a 4×4 black-and-white photobooth grid of the same person in all 16 frames. Use the uploaded photo as the only identity reference and keep the exact face, skin tone, proportions, messy high bun with curtain bangs, and outfit consistent.
Expressions in order, left to right, top to bottom:
soft smile; hand binoculars; laughing eyes closed; bored cheek-on-hand; duck face; shocked hands-on-cheeks; tongue out; scrunched face one eye closed; puffy sad face; hands on head duck face; thinking hand-on-chin looking sideways; big laugh; peace sign duck face; playful claw roar; eyes-closed kiss face; one eye covered with subtle smile.
Style: professional studio photobooth, seamless white background, high-contrast monochrome, ultra-realistic DSLR, sharp focus, natural skin texture, head-and-shoulders framing, evenly spaced 4×4 square collage, minimal shadows, premium editorial quality.
Avoid: extra people, identity changes, different hair/outfit, duplicate expressions, blur, cropped face, bad hands/anatomy, text, logos, watermarks, or distorted frames.