Did you know that we have LinkedIn, Instagram and X accounts
that you can follow?
In this issue:
🤝In Partnership: Turn one draft into every post
🤿Deep Dive: Muse's newest model spent 24 hours optimizing a single GPU kernel
🤝Powered by: The faster way to prompt AI
🤿Deep Dive: Handoff claims to work on sites with no API
⚒Tool Snapshots: Tools for AI, no-code, and productivity
🖼AI Art: Examples of great and trending AI art
🤝IN PARTNERSHIP WITH SURE THING
One idea shouldn't take six rewrites to post.
Posting everywhere means rewriting one idea six times, so you post to one, or none. SureThing turns one idea into native posts for every platform.
🤿 DEEP DIVE
Meta's Muse Code Keeps Background Subagents On The Job All Session Long
Intelligence: Meta rolled out a terminal coding agent designed to work across entire repositories, backed by a model built specifically to sustain long, multi-step engineering work. The release includes a runtime built to survive crashes and a training approach aimed at making the model grade its own homework.
Image credit: Meta
Muse Code's background subagents stay active for the full session instead of spinning up fresh for each task, cutting down on repeated context gathering.
The runtime logs every model call, tool run, and edit, letting Muse Code restart from the exact point a crash occurred.
Muse Spark 1.2 was trained using its predecessor to generate difficult coding challenges and grade the results, creating a self-improving feedback loop.
In one test, Muse Spark 1.2 spent up to 24 hours and over 1,000 tool calls optimizing GPU kernels for Hopper hardware, improving on the baseline the whole way through.
🤝POWERED BY WISPR FLOW
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
🤿 DEEP DIVE
Hark Says Its Model Predicts Actions Instead of Tokens
Intelligence: Fresh off a $700 million round, Hark released its first public product, a browser agent meant to click and type its way through sites that were never built for automation. The company argues its approach fundamentally differs from how large language models work, though a partial demo leaves plenty unconfirmed.
Handoff reads both page structure and visual layout to decide where to click or type, according to Hark.
Hark says Handoff predicts the next action rather than the next token, treating clicks and keystrokes as its basic unit.
The current release runs on a post-trained model, with Hark planning full pretraining later this year.
In a demo, Handoff built a flower order from vague instructions like "florist's choice," though the clip cut off before the process finished.
Hark claims Handoff beats rivals like GPT-5.5 and Opus 4.8 on speed and cost, a claim still unverified ahead of its end-of-summer release.
⚒ TOOL SNAPSHOTS
Futuristic tools within AI, no-code, and productivity
☁️ Cloudflare OS - Brings your company's knowledge and tools into one place for AI agents.
💸 AI Spend Console by Rippling - Makes it easier to see where AI budgets are going and what's driving them.
💬 CopilotKit Channels SDK - Helps your agents feel at home wherever your team already chats.
🛠️ Keystroke - Cuts down the work between an idea for an agent and putting it to use.
📝 Aveiro - Keeps publishing across channels from turning into a copy-and-paste routine.
🖼 AI ART
Examples of great and trending AI art
Images by Lazy_Lime419
ℹ️ ABOUT US
The Intelligent Worker helps you to be more productive at work with AI, automation, no-code, and other technologies.
We like real, practical, and tangible use-cases and hate hand-wavy, theoretical, and abstract concepts that don’t drive real-world outcomes.
Our mission is to empower individuals, boost their productivity, and future-proof their careers.
We read all your comments - please provide your feedback!
Did you like today's email?
What more do you want to see in this newsletter?
