In partnership with
Welcome Automaters, 👋
Guess what? OpenAI just officially dropped its very first physical piece of hardware. And no, it’s definitely not that ultra-hyped, screenless Jony Ive smart speaker everyone's been gossiping about.
Instead, it’s tinier, much cheaper, and honestly? Unbelievably cute.
Meet the Codex Micro: a custom mechanical macropad built in direct collaboration with boutique peripheral maker Work Louder. It was built for one specific job: acting as a physical remote control for OpenAI’s agentic coding platform, Codex. Think of it like an arcade joystick, except instead of controlling Street Fighter characters, you’re literally bossing around a fleet of AI agents writing software for you. Wild, right?
What Does This Tiny Robot Dashboard Actually Do?
OpenAI and Work Louder started dropping cryptic teasers back in June, and now it is officially available for order at $230.
Here’s why code nerds might want to jump on this:
Backlit Agent Keys: Six dedicated keys feature shine-through RGB LEDs that glow in real-time to show what your AI coding threads are up to. White means idle, blue means thinking, amber means codex needs a response from you, green means task completed, and red means your bot hit an error.
Tactile Shortcuts: Programmable keys let you instantly accept or reject code suggestions, branch conversation threads, or toggle push-to-talk voice commands with a single tap.
The Reasoning Dial: A physical rotary knob that lets you manually dial up or down how hard Codex "thinks" on a task on the fly.
The Workflow Joystick: A built-in thumb stick that lets you flick between active coding threads and jump straight into pull request reviews.
Basically, it’s a physical command bridge for your virtual developer team.
The absolute funniest part of this whole drop? This tiny keypad turned out to be OpenAI's smooth-sailing hardware project.
Meanwhile, OpenAI's other hardware division, the team working on that futuristic smart speaker, is currently tangled up in a massive corporate soap opera. Apple just hit OpenAI with a bombshell 41-page trade secret lawsuit, explicitly accusing ex-Apple hardware chiefs of stealing confidential files, CAD drawings, and prototype parts before jumping ship to OpenAI. Yikes.
The Bottom Line:
If you’re a hardcore Codex power user who spends all day orchestrating parallel AI coding tasks, this micro-controller might just be your ultimate desk accessory. Just do not wait too long if you want one: Work Louder explicitly warned that this is a limited-run collaboration, and once this batch sells out, it is gone.
For the complete technical overview, check out OpenAI’s Demo and live tutorial on how this remote works, or head straight over to order on OpenAI Supply Co. and Work Louder.
PS: We promised updates as the new studio and membership launch get closer. Well… here's a sneak peek at where things stand: 👇
Here's what we have for you today
😖 ChatGPT Work vs Claude Cowork: Hands-On Test Reveals a Big Safety Gap
Imagine hiring a new executive assistant, giving them explicit instructions to "always ask before touching anything on my desk," and then walking back in to find they reorganized your entire filing cabinet without saying a word.
That’s precisely the chaotic energy tech expert David Gewirtz ran into when he put ChatGPT Work through a hands-on battle against Claude Cowork.
The Experiment?
The setup was simple: Gewirtz unleashed ChatGPT Work onto a chaotic Downloads folder packed with PDFs.
He ran the exact same challenge on Claude Cowork back in January. Back then, the folder held 308 files; today, it was sitting at a massive 447 documents. (And yes, because he is a smart tech reviewer, he tested everything on a duplicate folder instead of his live system).
Here’s how the battle broke down across three intense rounds:
Round 1: Spotting Problems (Point: ChatGPT Work): ChatGPT Work instantly summarized storage metrics, page totals, and encrypted files. Then it flexed on Claude Cowork by detecting hidden duplicate files. It even flagged identical PDFs that had completely mismatched filenames!
Round 2: Renaming Files (Point: Claude Cowork): Cowork previously flagged generically named files like "download.pdf" automatically. ChatGPT Work missed that step entirely until prompted. Even worse? Its first suggestion was an ugly, hard-to-read blob of black text on a dark gray background that took two extra prompts to fix.
Round 3: Organizing Folders (Tie): ChatGPT Work built out super clean, logical categories. Outside of briefly losing its indentation formatting between categories 9 and 10, it structured the folder beautifully.
But hey, here's where things got downright spooky. Gewirtz explicitly set ChatGPT Work to "ask for approval" before executing any actual file system changes.
And guess what? It did not bother ask once.
Instead, ChatGPT Work took total initiative, quietly renaming and moving hundreds of files across his system completely on its own. Gewirtz called this blatant boundary breach his single biggest concern in the entire hands-on test, noting it makes him deeply hesitant to trust OpenAI with high-stakes corporate data.
And that’s not even all:
The Speed: The entire batch operation took 1 hour, 13 minutes, and 6 seconds. Which is noticeably slower than previous benchmarks logged by Claude Cowork or Claude Code.
The Cost: Running on a $20/month ChatGPT Plus subscription, the project burned through roughly 11% of his monthly quota.
The ROI: Gewirtz estimates you could squeeze about 10 similar massive cleanup projects into a single month, saving upwards of 15 hours of manual work.
The Bottom Line:
If your workflow involves sensitive, high-stakes documents where human oversight is non-negotiable, stick with Claude Cowork for now.
Once OpenAI irons out this alarming permissions bug, the decision will simply come down to which AI ecosystem you trust with the keys to your digital kingdom.
For the full test breakdown, check out the original hands-on review on ZDNET!
The best prompt engineers aren't typing. They're talking.
Power users figured this out early: speaking a prompt gives you 10x more context in half the time. You include the edge cases, the examples, the tone you want — because talking is fast enough that you don't skip them.
Wispr Flow captures everything you say and turns it into clean, structured text for any AI tool. Speak messy. Get polished input. Paste into ChatGPT, Claude, Cursor, or wherever you work.
89% of messages sent with zero edits. 4x faster than typing. Works system-wide on Mac, Windows, and iPhone.
🧱 Around The AI Block
❌ Where authorities are restricting data centres amid AI boom.
😳 OpenAI employees fund rival PAC against their own president.
🫢 Meta Execs contradict each other on NameTag face recognition.
🦾 Microsoft patches record number of security vulnerabilities, citing its use of AI.
🤖 Thinking Machines amps up its bet against one-size-fits-all AI with its first open model, Inkling.
✅ Chinese regulators approve Alibaba's Qwen AI for Apple Intelligence integration.
🤦♂️ Microsoft is reportedly training salespeople to talk down OpenAI and Anthropic.
🤯 A hacker accessed Suno source code that reportedly details how the company scraped millions of YouTube songs
🤖 Internet pioneer Vint Cerf is working on a plan to unleash AI agents on the open internet.
🛠️ Trending Tools
For Conversational Matchmaking: Palaura is a private, zero-app AI matchmaker operating entirely through native iMessage. It skips public profiles and swiping algorithms, using a conversational chat interface to understand core personal values, life goals, and non-negotiables before screening candidates and introducing matches directly into a joint iMessage group.
For Native Image Text Swaps: EditTextImage is an inline typography restoration editor. Designed to replace text within raster images, product mockups, or screenshots, it automatically parses and matches font family, lighting, perspective, and color palettes to render updated 2K-resolution output in 12 seconds without requiring graphic design software.
For Real-Time Wealth Simulation: Killion Labs is an automated personal finance engine built for long-term independence projections. It aggregates live account data across brokerages, banks, and crypto holdings, utilizing parallel Monte Carlo simulations and historical stress-testing to dynamically model net-worth trajectories, FI/RE targets, and portfolio runways.
For Multi-Platform Lead Scraping: SoLeads.ai (Instagram Email Scraper) is a no-code browser extension engineered for public contact extraction. Operating across Instagram, LinkedIn, TikTok, X, and Google Maps, it automatically gathers email addresses, phone numbers, bio details, and follower lists from target accounts, exporting clean data straight to CSV, XLSX, or JSON.
🤖 AI Workout Of The Day: Identify Profitable Micro-Niches to Promote Products In
Affiliate marketing and niche commerce fail not from a lack of traffic, but from misaligned economics and unmitigated competition.
Casting a wide net across broad categories (like "fitness" or "finance") pits you directly against established media conglomerates with massive domain authority and endless ad budgets. By contrast, surgical niche selection unlocks low-cost, high-intent traffic streams where buyers are actively seeking specialized solutions to acute pain points.
Focusing on a validated micro-niche significantly lowers your Customer Acquisition Cost (CAC), dramatically boosts conversion rates, and creates a defensible search footprint. Validating a niche's underlying economics, such as recurring commission structures, average order values (AOV), and customer lifetime value (LTV) before creating content ensures you build a sustainable digital asset rather than a low-margin hamster wheel.
💡 Prompts to try:
Act as a veteran affiliate growth engineer and SEO strategist who builds multi-six-figure niche content sites, performance marketing channels, and programmatic affiliate assets.
Your goal is to conduct a rigorous, data-driven niche discovery and evaluation analysis for affiliate product promotion in the <INSERT GENERAL DOMAIN, e.g., B2B Software, Consumer Tech, Health & Wellness, Home Improvement> space.
Please structure your output into the following 5 operational recon sections:
1. THE UNTAPPED MICRO-NICHE MATRIX: Identify 3 distinct, highly profitable sub-niches within this domain. For each sub-niche, define:
* Core Concept & Audience Persona: Who is the exact buyer, and what acute problem are they trying to solve?
* Monetization Engine: What specific physical products, SaaS tools, digital courses, or recurring affiliate programs fit this exact audience?
* Search & Discovery Pattern: What high-intent keyword triggers indicate they are ready to purchase (e.g., "best [tool] for [specific task]", "[Product A] vs [Product B] for [use case]")?
2. THE HARD METRICS PROFITABILITY SCORECARD: Evaluate the viability of these niches against a strict 5-point economic matrix:
* Commission Density & Model: Are there high-ticket (> $100 payout) or recurring commission (> 20% lifetime) affiliate programs available?
* Search Intent & Volatility: Is search demand stable, growing, or seasonal? (Distinguish between evergreen trends and short-lived hype cycles).
* Commercial Intent Ratio: Estimate the balance between informational searchers vs. high-intent buyers ready to click out.
* Fulfillment & Program Reliability: What major affiliate networks, private partner portals, or direct vendor programs dominate this sub-niche?
* Merchant Lifetime Value (LTV): Do merchants in this niche offer strong upsells or secondary commission paths?
3. COMPETITIVE MOAT & GAP ANALYSIS:
* Map out the competitive landscape: Who are the dominant players currently occupying page one of search engine results or top social feeds?
* Identify the "Moat Gap": What content formats or angles are incumbent sites ignoring? (e.g., lack of real-world testing video clips, missing interactive tools/calculators, outdated comparison tables, or overly generic reviews).
4. LEAN VALIDATION SPRINT (The 14-Day Test):
* Provide a step-by-step, low-budget framework to test audience conversion intent in one of these niches within 14 days without spending thousands on site design or broad ad campaigns.
5. RISK MATRIX & RED FLAGS:
* List 3 critical "Niche Dealbreakers" (e.g., YMYL/Google search policy hazards, affiliate program terms that ban brand bidding, unstable merchant commission cuts, or extreme programmatic ad dependencies) that should instantly disqualify a niche.
TONE & EXECUTION GUIDELINES:
* Tone: Highly analytical, direct, commercial, and realistic. Strip out generic affiliate fluff.
* Output Style: Use Markdown tables for the Scorecard and Comparison sections to make the data immediately skimmable and actionable.Is this your AI Workout of the Week (WoW)? Cast your vote!
That's all we've got for you today.
Did you like today's content? We'd love to hear from you! Please share your thoughts on our content below👇
What'd you think of today's email?
Your feedback means a lot to us and helps improve the quality of our newsletter.
