In partnership with
The integrated coworker for AI native teams
Empower your team to do their best work with Adapt, the integrated coworker that works alongside your team in Slack and deeply understands your business.
Here’s how Adapt is different
Set up takes minutes: connect your tools, add to Slack, and it’s right there for anyone to tag @Adapt for help
Does real, high-ROI work: automates work on a schedule; builds internal tools with live data; and does complex, multi-tool tasks on demand
Learns your business as you work, becomes your company brain
Uses the best AI model for the task, not tied to a single provider
SOC2 Type II, RBAC, and support for personal and company-wide integrations
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
Hims Founder Used AI to Make 50 Ads a Month for What 5 Used to Cost, And Fired His Ad Agency
TLDR: A Hims cofounder dropped the outside ad agency costing his pet telehealth company $50,000 a month, built a small in-house team that runs on AI tools, and now makes ten times more ads while paying 20% less for every new customer.
The Story:
Joe Spector helped build Hims & Hers before starting Dutch, an online vet service, back in 2021. Members pay $100 a year for video visits covering up to five pets. Late last year, Dutch was handing an outside agency about $50,000 a month for five pieces of creative work. That's roughly $10,000 apiece. Spector told Fortune it was expensive and limiting, because so much of marketing a new product is just testing to see what connects. So he cut the agency loose and put a small in-house team in its place, one that uses AI tools to make everything from still ads to animated and live-action commercials. Dutch now turns out about 50 pieces a month, and Spector says buying that much work from an agency would run close to half a million dollars.
The numbers moved fast. Comparing January 2026, before the new team started, with April, Dutch cut what it pays to land a new customer by 20%. Its audience grew 20% month over month from February to March. The slice of its ad budget going to Meta climbed from about 5% to roughly a quarter. Asked about it, Spector pointed away from the software and toward his staff, saying "I shouldn't take the credit" and giving it to the marketing folks instead.
Its Significance:
Read the headline and it sounds like AI took a bunch of people's jobs. Look closer and something else happened. Spector didn't fire his marketing staff. He fired the vendor, hired people in-house, and gave them tools that let a small group do what used to take a whole company. That's a very interesting trend if you work in anything creative. The risk isn't only a machine replacing you. Companies are also deciding they don't need to rent an outside team anymore, because four or five people with good tools can cover it. And Spector says the savings weren't even the main goal. What he wanted was the ability to test more ideas, faster, until he found the ones pet owners actually respond to.
QUICK TAKES
The story: A school district in Salamanca, New York, spent $57,590 on a humanoid robot named Sally plus an AI teaching assistant, planning to put it in 11th and 12th grade coding and robotics classes this fall. Teachers called the plan inappropriate and said it strips the human part out of the job, some Seneca Nation residents called the choice of their district culturally tone deaf, and the district paused the whole thing on July 24 while it sorts out student data privacy agreements.
Your takeaway: The district serves about 1,300 students. What stopped the rollout was an unfinished agreement over who gets to hold kids' data, and that question is going to hit a lot more school districts this fall, robot or no robot.
AI redesigned gene-editing tools so they stop cutting the wrong DNA
The story: Researchers at Peking University and East China Normal University built a system called ContactSeek that uses Google DeepMind's AlphaFold3 to predict exactly where a gene-editing protein touches DNA, then flags the spots that make it grab the wrong sequence. Their best redesigned editor needed only two changes and beat several existing high-accuracy editors, according to the paper published in Nature on July 22.
Your takeaway: Gene editing's oldest problem is the off-target cut, where the tool edits a stretch of DNA that looks like the target but isn't. Fewer wrong cuts means safer gene therapies later on. This one ran in cells, not in people, so real treatments are still a long way off.
US lab initiative targets 5x faster engineering with next-gen supercomputing
The story: The Energy Department picked a company called Rescale for early funding under the Genesis Mission, its plan to tie 17 national labs and roughly 40,000 scientists into one research setup. Rescale will work with the Berkeley, Livermore, and Oak Ridge labs to put AI agents on top of heavy simulation software, so the agents pick the right tool, set up the inputs, watch the run, and read back the results.
Your takeaway: The labs write some of the best engineering simulation code anywhere, and almost nobody outside the labs can run it without a specialist on staff. Cutting the time and know-how needed by more than five times would put that software within reach of regular US manufacturers, not just people with a PhD and a supercomputer.
TOOLS ON OUR RADAR
📖 whoami.wiki Free and Open Source: A brilliant personal encyclopedia that uses artificial intelligence agents to document your family history and personal stories by organizing your loose photographs and memories into a private Wikipedia. (Alternative to Ancestry)
🌳 KnowTree Freemium: A unique conversation mapping tool that transforms your flat chat history into branching navigable trees allowing you to visually track and compare different paths of thought across multiple models simultaneously.
📄 Storydoc Freemium: An interactive business document creator that uses artificial intelligence to transform your static decks and reports into engaging web based experiences complete with embedded forms and real time engagement analytics.
🎓 OpenMAIC Free and Open Source: An innovative educational platform that transforms any document or topic into a full multimedia classroom experience with artificial intelligence teachers and interactive classmates.(Drop the github link into your favorite AI and tell it to access that way.)
TRENDING
Spider-inspired four-legged robot boat could pull people out of the water - Researchers in Shenzhen and at Stevens Institute of Technology built QuadBoat, a four-legged robot boat modeled on fishing spiders. Most rescue boats can only reach a person in the water. This one is built to lift them out.
The running list: major tech layoffs in 2026 where employers cited AI - TechCrunch is keeping a running tally of big tech companies that cut jobs this year and named AI as a reason, from Oracle's 21,000 down to Amazon's 16,000. Most of them posted record revenue in the same stretch.
Librarians are hosting viral 'Avoiding AI' workshops for people who are fed up with Big Tech - A Philadelphia librarian's hour-long class on switching off Apple Intelligence and Gemini drew so many signups he had to add a second session. The library's post about it pulled over 2,000 likes, against the few dozen its posts normally get.
DeepSeek tells prospective investors of funding pause - The Chinese AI company told backers it's suspending its second funding round for now, per Bloomberg. It had been aiming at a valuation near $74 billion after raising about $7.4 billion in its first round, and it may pick the deal back up later.
AI companies are buying rare books, scanning them, then destroying them - Booksellers say AI labs are buying used and out-of-print books in bulk, slicing off the spines to feed the pages through fast scanners, and pulping what's left. A US court ruled in the Anthropic case that buying, scanning, and destroying a book you own counts as fair use, which is a big part of why it keeps happening.
7 AI technologies making military drone swarms smarter and deadlier - A rundown of what's actually running inside a drone swarm, including the decentralized algorithms that let dozens of drones hold formation, split up jobs, and keep working after some get shot down.
The first way to trade directly inside Claude and ChatGPT
Superintelligence used to be locked inside billion-dollar quant firms whose algorithms quietly took advantage of everyone else.
Co-Invest puts it right in your chat window. Analyze markets, manage risk, and execute trades, all inside Claude and ChatGPT.
The institutions built the game, Co-Invest gives you a way to beat them.
TRY THIS PROMPT (copy and paste into Claude, ChatGPT, or Gemini)
📸 Photograph what you actually own. Mix it live on a digital mannequin, then get an honest AI read on what those exact pieces give you.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Capsule Wardrobe Builder — photograph your clothing pieces, mix them on an interactive digital mannequin board, then get an AI capsule-wardrobe analysis of the real pieces via the vision API. Persist to localStorage key 'capsule_wardrobe_v1' (analysis text only, never the photos).
Aesthetic: near-black fashion-editorial (#0d0d10), hot pink/rose (#e0447a) primary with a soft pink glow top-left. Bodoni Moda (dramatic high-contrast serif) for headings/combo names, Inter for body, JetBrains Mono for labels. A minimalist faceless croquis-style mannequin drawn in muted gray SVG line art as the board's background.
Upload: a tap-to-add multi-photo zone (accept="image/*" multiple, cap at 12). Each uploaded photo becomes a small chip showing a thumbnail, a removable ✕, an editable label text input (e.g. "navy blazer"), and a category select (top/bottom/outerwear/shoes/accessory).
Mannequin board: an SVG croquis silhouette as a faint background layer inside a fixed-aspect-ratio container, with 5 absolutely-positioned dashed-border image slots layered over the body regions (outerwear widest behind, top layered in front of it, bottom below, shoes at the feet, a small accessory slot near the collar). Each slot shows the currently selected item from that category (object-fit: cover) with small ‹ › circular arrow buttons on its edges to cycle through all items tagged to that category, and a small caption strip showing "label · i/n". Below the board: a Randomize button (picks a random index per populated slot), a look-name text input, and a Save This Look button that appends the current combo (category → label) to an in-memory (not persisted) "today's looks" list.
Analyze: a separate "Analyze My Capsule" button sends ALL uploaded photos as an array of {type:'image', source:{type:'base64', media_type, data}} blocks followed by one text block listing each item's category and label, in a single Messages API call (no web_search needed). When parsing the response, filter content blocks for any with a text property and concatenate before stripping JSON fences.
System instructions: expert wardrobe stylist analyzing the ACTUAL photographed pieces — real colors, patterns, fabric weight, formality, silhouette — never inventing items not shown. No em dashes. Return raw JSON: style_read (2-3 sentences grounded in what's actually visible), combo_estimate (short phrase, rough combination count), versatility_tags (2-3 short tags), top_combos (exactly 3: name, pieces using the given labels, why referencing actual observed details), gap (1-2 sentences on the one piece type that would unlock the most new combinations), quirk_note (optional standout observation).
Render: gradient "the read" style card with combo-estimate and versatility tag chips. Three combo cards (name, mono piece list, italic why). A pink "add this next" gap card. A small italic quirk note. Archive of past capsule analyses keyed by style-read snippet + piece count (photos never persisted).What this does: Upload photos of your individual clothing pieces, tag each as top, bottom, outerwear, shoes, or an extra, and give it a quick label. Your photos snap onto a faceless mannequin board with five zones you cycle through with arrow buttons, so you can flip through real combinations of your real clothes. Randomize for a surprise pairing, or save a favorite for the session. Then run the Analyze button, which sends the actual photos to a vision model for a capsule read: the style vibe it sees, a rough count of solid outfit combinations, three named combos using your labeled pieces with the actual color and style details behind each, and the one type of piece that would unlock the most new outfits if added. The mannequin board is a visual mixer using your real photos, not an AI-generated render, since that's not something a portable file can do.
What this looks like:
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Redesign a gene-editing protein so it locks onto the right stretch of DNA and skips the near-matches, using only two changes to the protein itself.
❌ Still Can't: Show those edits are safe in a living person. The work ran in cells, not patients.
✅ AI Can Now: Help a small in-house team produce about 50 finished ads a month, including animated and live-action video.
❌ Still Can't: Tell you which of those ads will land. Dutch still has to run them and measure what happens.
RECOMMENDED LISTENING/READING/WATCHING
Salvation by Peter F. Hamilton - Book
A 2018 sprawling near-future thriller, first of Hamilton's Salvation Sequence, about an alien starship discovered on a distant colony world, whose sudden reappearance forces a small team of specialists to investigate what the aliens want. The narrative alternates between their present-day investigation and, thousands of years later, human survivors on a distant world planning a counterattack, all tied together by AI-augmented memory, quantum entanglement communications, and post-human intelligences.
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.
