In partnership with
Speak messy. Prompt clean.
Go on tangents. Change your mind mid-sentence. Say "um" twelve times. Wispr Flow doesn't care — it takes everything you say, strips the filler, and gives you clean, structured text ready to paste into any AI tool.
The result: prompts with the full context your AI tools need to give you useful answers. Not the abbreviated version you'd type because typing is slow.
Works inside ChatGPT, Claude, Cursor, and every app on your screen. Millions of users worldwide, including teams at OpenAI, Vercel, and Clay.
Beginners in AI
Good morning and thank you for joining us again!
Welcome to this daily edition of Beginners in AI, where we explore the latest trends, tools, and news in the world of AI and the tech that surrounds it. Like all editions, this is human curated and edited, and published with the intention of making AI news and technology more accessible to everyone.
THE FRONT PAGE
AI Agents Closer to Permanent Memory. Here's What Changes
TLDR: A new tool called MemWal gives AI agents the long-term memory they've been missing, fixing one of the biggest reasons AI assistants forget your projects, preferences, and context the moment a session ends.
The Story:
Walrus, the company behind a network for storing data, just launched MemWal, a kit for developers that gives AI agents real long-term memory. Until now, most AI agents start every conversation from scratch. You teach them your preferences on Monday, and by Wednesday they've forgotten. MemWal stores those memories in encrypted form, lets you carry them between different AI tools, and lets you verify that nothing has been changed behind your back. The launch comes the same week another company called Engramme said its goal is to give humans "infinite memory" through AI, recalling every person you've met and every conversation you've had. Engramme was started by a Harvard neuroscience professor who has spent twenty years studying how the brain stores memories. Two different companies, both betting that memory is the next big problem to solve in AI.
Its Significance:
Memory is the reason AI assistants feel useful for one task and useless the next. If your AI remembers your job, your writing style, the names of your kids, and the project you've been working on for six months, it can actually help you. If it forgets every time, you spend half the conversation re-explaining yourself. The flip side matters too. We wrote about Engramme the other day and the worry isn't dystopian, it's quieter. The act of trying to remember a name or a moment is what builds the memory in the first place. If AI does all the remembering for us, that mental muscle can fade, the way spatial skills decline in heavy GPS users. Pick tools that help you remember better. Be careful with ones that offer to do the remembering for you.
QUICK TAKES
The story: France's Mistral released Mistral Medium 3.5 on April 29, a 128-billion-parameter model with new coding agents and a Work Mode for handling email and research. The problem: it costs about $1.50 per million input tokens and $7.50 per million output tokens, while Chinese open-source models like Qwen 3.6 score about the same on coding tests at one-fifth the size and a fraction of the price.
Your takeaway: Europe still has only one major AI lab in the open-source race, and it just shipped something that costs multiples more than the Chinese competition without beating them on the benchmarks. If you pick AI tools for your work, the cheap, fast options are mostly coming from China right now. That's a real shift from two years ago.
The story: OpenAI launched Advanced Account Security on Thursday, an opt-in setting aimed at journalists, dissidents, researchers, and anyone whose ChatGPT account has become a place where they store sensitive work. It requires a passkey or a physical security key (like a YubiKey) to log in, removes email and SMS recovery, and automatically excludes your conversations from training data.
Your takeaway: ChatGPT accounts now hold a lot more than chat history. They hold your work documents, your business ideas, sometimes your medical questions. If your account is locked to a phone number a hacker could SIM-swap, that's a real risk. The catch: if you lose your security keys, OpenAI can't get you back in. Worth turning on if your account holds anything you'd hate to lose.
The story: Canonical, the company behind Ubuntu (the most popular version of Linux), posted plans last Sunday to add AI features throughout 2026. Many Ubuntu users had switched to Linux specifically to escape the AI features Microsoft kept forcing into Windows, and the backlash was instant. Two days later Canonical clarified: features will be opt-in starting with Ubuntu 26.10, local processing is the default, and any AI piece can be uninstalled.
Your takeaway: If you've been thinking about leaving Windows because of the AI creep, Linux is still mostly safe. Ubuntu's plan keeps your data on your own machine by default, and you can remove the AI bits if you want. But the era of an operating system that ignores AI completely is probably ending. Even Linux is getting some.
TOOLS ON OUR RADAR
📸 DigiKam Free and Open Source: A remarkably powerful photo management program that makes it easy to organize edit and search through thousands of your digital family pictures using advanced facial recognition and tagging.
🗜️ PeaZip Free and Open Source: A secure and beautifully designed file utility that replaces annoying paid programs by easily opening extracting and creating compressed folders so you can share multiple large files at once.
🎬 OpenToonz Free and Open Source: A powerful two dimensional animation studio famously used by Studio Ghibli that provides traditional and digital artists with all the tools needed to create stunning animated films.
⚔️ Habitica Freemium: A completely unique task manager that turns your daily chores and habits into a retro role playing game where you earn experience points and virtual gold for staying productive.
TRENDING
Amazon's New AI Podcasts Want to Sell You Diaper Rash Cream — Amazon launched AI hosts that record fake "podcast" segments hyping individual products. A Business Insider reporter got one for adult diaper rash cream. The same feature has gushed over fake dog poop. We've apparently spent billions on AI to reinvent the late-night infomercial.
Perplexity Opens Computer to All Pro and Enterprise Users — Perplexity Computer, the cloud-based AI agent that runs 20 different models to handle research, coding, and design from a single prompt, is now available to every Pro and Enterprise Max subscriber, no longer just the $200/month Max tier. Connects to Snowflake, Salesforce, HubSpot and 400 other apps.
Zuckerberg Says Meta Layoffs Are About AI Spending and More Cuts May Come — At a town hall Thursday, Mark Zuckerberg told Meta employees that the planned May 20 layoffs (about 10% of staff) are happening because the company is pouring money into AI compute and has to cut somewhere else. He won't say no to more cuts later this year.
China's New Cars Are Picking Sides in the Doubao vs. Qwen War — At the Beijing Auto Show, Alibaba said its Qwen AI is going into BYD and Volkswagen-China cars. Audi's new electric SUV picked ByteDance's Doubao. Cadillac's new China model also picked Doubao. As EV prices crash, automakers are using AI features as the new way to stand out.
Anthropic Launches Claude Security to Find Bugs in Your Code — Claude Security, powered by Opus 4.7, is now in public beta for all Claude Enterprise customers. It scans entire codebases, finds security flaws (including ones existing tools missed for years), and writes patches you can apply through Claude Code. Hundreds of organizations tested it during the preview.
Spotify Rolls Out a Green Checkmark to Prove Artists Are Real People — Spotify launched "Verified by Spotify" badges Thursday after Sony Music asked the platform to remove 135,000 AI-generated songs impersonating its artists. AI-only profiles aren't eligible. About 99% of artists listeners actually search for will be verified at launch, mostly independent artists.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
🪫Answer 8 honest questions. Get your burnout score, dimension breakdown, and a personalized recovery prescription.
Build a single-file HTML app using vanilla HTML, CSS, JS, and one Claude API call. Create Burnout Barometer — an 8-question burnout assessment with scoring, dimension breakdown, and recovery prescription. Save all results to localStorage key 'burnout_barometer_v1'.
Aesthetic: off-white paper background (#f5f2ec), stark black typography, red (#c0392b) as the danger accent. Typography: Syne 800 for headings with tight letter-spacing, Lora italic for prescriptions, Syne Mono for labels. Clinical but human.
Eight questions covering: Exhaustion, Detachment, Efficacy, Recovery (inverted), Autonomy (inverted), Resentment, Meaning (inverted), Outlook (inverted). Each has a 1–5 scale with descriptive endpoints. Selected buttons color-code: red for 4–5, amber for 3, green for 1–2. A progress bar fills as questions are answered. Submit button enables only when all 8 are answered.
Calculate a 0–100 burnout score. Invert the appropriate dimensions before scoring. Send score and all 8 dimension values to Claude API requesting raw JSON: label, description, optional warning (if score ≥ 70), immediate actions array, structural changes array, reframe string.
Render: animated count-up score display with color (red/amber/green), 8 dimension bars with color-coded fills, prescription sections for this week and structural changes, a reframe quote at the bottom. Store each assessment with date, score, label, and parsed data. Show past assessments as a history list with color-coded scores. Make it work in a single HTML file.What this does:
Eight questions covering exhaustion, detachment, efficacy, recovery, autonomy, resentment, meaning, and outlook. Each answered on a 1–5 scale. Claude scores you 0–100, names your burnout stage, breaks down all eight dimensions as progress bars, flags an urgent note if your score crosses 70, and delivers two sets of prescriptions: what to do this week and what structural changes to make longer-term. It also ends with a single honest reframe — something true and helpful about being at that stage. Every assessment saves to localStorage with the date and score, so you can track whether things are actually getting better.
What this looks like:
Do your searches always hit dead ends?
Tired of long paragraphs from your AI chatbot when you just need to make a quick decision?
heywa answers your questions with visually curated stories instead of walls of text.
Ask heywa a question, refine by what you need and tap through.
Decisions just got easier.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Keep long-term memory across sessions, so your agent can pick up a project where you left off weeks ago
❌ Still Can't: Remember things securely without storing that memory somewhere a third party (or a hacker) might be able to read
✅ AI Can Now: Scan an entire codebase and find security flaws that human experts missed for years
❌ Still Can't: Tell, on its own, whether the song you're listening to was made by a real person or a synthetic profile farming streams
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
In a parallel present, lifelike humanoid robots called Synths serve as household servants, workers, and companions, but some of them might be developing consciousness. The first episode introduces a family who buys a Synth that clearly has more going on than the manual suggests. It's a slow-burn British drama that asks uncomfortable questions about servitude and personhood.
Thank you for reading. We’re all beginners in something. With that in mind, your questions and feedback are always welcome and I read every single email!
-James
By the way, this is the link if you liked the content and want to share with a friend.
Some * designated product links may be affiliate or referral links. As an Amazon Associate, I earn from qualifying purchases. This helps support the newsletter at no extra cost to you and Amazon makes a tiny hair less.
