In partnership with
Exploring AI Voice With SuperBloom
Scaling a campaign globally means finding a voice that resonates everywhere—without losing the overall message. For Deel's "Feeling of Deeling" campaign, agency SuperBloom needed exactly that: a consistent brand voice across markets, deployed fast, without sacrificing quality or consent. They built it with Branded AI Voice, powered by real, professional talent.
SuperBloom and Voices break down how the campaign came to life in this on-demand video session—from strategic talent selection through seamless production workflows and global scale, to the governance that future-proofs an audio strategy built to last. You'll hear directly from the team on how they made this happen, plus their advice if you're looking to explore AI voice for your brand.
If you're a marketing executive, agency creative, or brand leader mapping campaigns in international markets, watch this on-demand session to get a real playbook, not a hypothetical—lessons any creative team can apply to their own global rollout.
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
An AI Drug Made Patients Biologically Younger, and It Was Built for Lungs
TLDR: A drug designed by AI made sick patients look three to six years younger on six separate blood tests of biological age, according to a new study in Nature Biotechnology.
The Story:
Insilico Medicine built a drug called rentosertib without starting from a human hunch. Its AI picked the protein to attack, a target called TNIK, and a second AI system designed the molecule to hit it. Target to drug candidate took about 18 months. The drug was meant for idiopathic pulmonary fibrosis, a lung disease that scars lung tissue and kills most patients within three to four years.
Researchers went back to blood drawn during a 42-patient trial and measured 2,841 proteins. Then they ran six "aging clocks," separate models built by teams at Harvard, Peking University and elsewhere, that guess your biological age from your blood. All six said the treated patients looked younger. The strongest reading came at week 4 in the 30 mg twice-daily group: three to four years of reversal, and up to six years on one clock. Placebo patients barely moved.
Its Significance:
If the shape of this sounds familiar, it should. Last spring, Sydney data engineer Paul Conyngham used ChatGPT and AlphaFold to help design a custom mRNA cancer vaccine for his dying dog Rosie, paid $3,000 to sequence her tumor, and got scientists at two Australian universities to build it. Her biggest tumor shrank by roughly 75%, though her vets never isolated how much of that came from the vaccine, and oncologists told him to slow down. Same method, wildly different budgets. Read a tumor or a protein profile as data, let AI pick the target, build the molecule to match. Insilico did it with a pharma company and a Phase III trial in China. He did it with a laptop and a credit card. You can't buy either one, and the FDA still doesn't count aging as a disease you can get a drug approved to treat.
QUICK TAKES
The story: Terence Tao, the UCLA professor who won the Fields Medal in 2006, posted on Mathstodon that AI labs are using up the supply of good unsolved math problems faster than mathematicians can find new ones. In May an OpenAI model disproved the Erdős unit-distance conjecture, an 80-year-old question, and outside mathematicians including Tim Gowers checked the work.
Your takeaway: Tao wants certain problems marked "analysis-required," so a bare AI answer with no explanation counts for little. Right answers and explanations are two different jobs. AI got very good at the first one first.
The story: PISA tested more than 760,000 fifteen-year-olds across 91 countries. Teens who almost never used AI to draft writing scored 509 in science, while daily users scored 481, a 28-point gap that works out to roughly a year and a half of school.
Your takeaway: The exceptions are the interesting part. Students who used AI once or twice a week to help them learn scored highest of any group, and daily users did 13 points better when their classes made them pick apart AI answers. How it gets used matters more than how often.
The story: The NASA-IBM Lunar Foundation Model trained on about 2 million image tiles from the Lunar Reconnaissance Orbiter, plus data from GRAIL and Japan's SELENE. NASA says it beats the older SwinV2-B system by up to 23% at spotting craters, volcanic formations and possible ice deposits.
Your takeaway: When a Falcon 9 rocket crashed into the moon, the team fed in the image and the model found the fresh crater on its first try, even though it overlapped an older one. The model sits on Hugging Face and the code sits on GitHub, so a small university lab can pick it up tomorrow.
TOOLS ON OUR RADAR
☁️ Sync.com Freemium: Store and share files with end-to-end encryption included on every tier, even the 5GB free plan.
🎤 Prezi Freemium: Build zooming, motion-based presentations instead of static slides, the free Basic plan works but adds a watermark and no offline export.
🎧 Audacity Free and Open Source: Record and edit multi-track audio with noise reduction and effects, just hit version 4.0 with a rebuilt interface.
📝 Google Keep Free: Jot quick notes, checklists, and voice memos that sync instantly across devices, no upgrade tier exists, it's fully free as part of a Google account.
TRENDING
Google and NASA Built an AI That Finds Methane Leaks From Orbit — Google Research and NASA's Jet Propulsion Laboratory released MAPL-EMIT, trained on 3.6 million simulated methane plumes because no big real-world dataset existed. It finds about 50% more plumes than human analysts, roughly 23,000 extra worldwide, including leaks at 24 of the 25 largest-emitting landfills. The plume database is on Earth Engine and the model is on Kaggle.
Prime Video Is Reshaping Actors' Mouths to Match Dubbed Audio — Amazon turned on AI lip-syncing for the German series Maxton Hall, live worldwide on seasons 1 and 2, with season 3 arriving December 9. Human actors still record the dub; the AI and VFX work happens afterward, changing the pixels around the mouth. Amazon won't say which systems do it.
LG Denies a Viral Claim That 216 Million of Its TVs Are Spying — Gamers Nexus published a two-hour video with two security researchers claiming LG smart TVs log and upload data, scan home Wi-Fi networks and capture audio in standby. LG told Tom's Hardware the claims are not true, while confirming the network scanning and confirming its TVs listen locally for the "Hi LG" wake word. Tom's Hardware says it has verified neither side.
An AI Now Decides Where a Microscope Should Look Next — Oak Ridge National Laboratory built SimuScan, which runs an atomic force microscope through a fast wide sweep, picks out things like DNA assemblies and bacterial cells, ranks them, then sends the instrument back for close-ups. The trick was training on synthetic images with deliberate flaws in them. Perfect fake data made a model that fell apart on real samples.
Life Sciences HR Is Moving AI Past Scheduling and Screening — Grace Niwa, who runs global talent acquisition at Vertex, told BioSpace the goal is cutting administrative work so recruiting stops being reactive and starts shaping talent decisions. The pitch to recruiters is that their job turns advisory instead of transactional. Whether the headcount survives that shift is a separate question nobody at these conferences is answering.
TRY THIS PROMPT (copy and paste into Claude, Grok, ChatGPT, or Gemini)
🌱 Enter your location. Get your real USDA hardiness zone, today's actual temperature, and what's genuinely worth planting right now.
Build a single-file HTML app with vanilla HTML/CSS/JS. The Garden Zone & Planting Calendar — enter a location, get the real USDA hardiness zone via live search plus real live weather, with what to plant this month. No API key needed.
Aesthetic: dark soil-brown (#181410), fresh green (#6cb85c) primary with a green glow top-left, sun-yellow (#e0c25c) secondary for weather stats with a glow bottom-right. Inter for headings/stats, Newsreader serif italic for the monthly context note, JetBrains Mono for labels.
Form: single location text input (city or zip).
Technical:
1. Geocode via geocoding-api.open-meteo.com/v1/search?name={location}&count=1.
2. Call the Anthropic Messages API WITH tools:[{type:'web_search_20250305', name:'web_search'}] (works automatically), system prompt as a gardening researcher who finds the REAL official USDA Plant Hardiness Zone for the location via search (planthardiness.ars.usda.gov or reliable gardening references), never approximating or guessing a zone number, plus real region-appropriate planting-calendar guidance for the current month (passed in dynamically). Return raw JSON: zone, zone_temp_range (real typical average annual minimum for that zone), current_month_note (2-3 sentences), plant_now (4-5: name, type, note), hold_off (2-3: name, why), frost_note.
3. In parallel: fetch real live weather via api.open-meteo.com/v1/forecast?latitude={lat}&longitude={lon}¤t=temperature_2m&daily=temperature_2m_min&temperature_unit=fahrenheit&timezone=auto&forecast_days=1 — real current temp and tonight's forecast low.
Render: a gradient hero card with the real zone number (large) and its temperature range. A 2-stat row (current temp, tonight's low). An italic serif monthly-context note. A green "good to plant now" list (name, type tag, note) and an amber "hold off on" list (name, why). A frost-context footer note.What this does: Type a city or zip and it searches the live web for your real, official USDA Plant Hardiness Zone rather than approximating one from raw temperature data, since zone boundaries follow the actual published map, not a simple formula. Alongside that sits real live weather: the current temperature and tonight's forecast low, useful frost context for timing.
What this looks like:
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.
WHERE WE STAND(based on today’s news)
✅ AI Can Now: Pick a brand new protein target, design a molecule to hit it, and reach human trials in about 18 months.
❌ Still Can't: Show that molecule is slowing aging rather than fixing one sick organ, which is why 42 patients isn't enough and a healthy-volunteer trial is the next ask.
✅ AI Can Now: Spot methane plumes in satellite images that trained human analysts miss, roughly 50% more of them.
❌ Still Can't: Stop firing false alarms over rough terrain, so Google ships confidence tags and leaves the filtering to you.
FROM THE WEB
RECOMMENDED LISTENING/READING/WATCHING
Blood Music by Greg Bear - Book
A 1985 Nebula and Hugo nominated novel about a genetic engineer who, ordered to destroy his most dangerous research, injects it into his own bloodstream instead to smuggle it out of the lab, unaware that what he's carrying is about to start thinking for itself. Bear is often called the natural heir to Arthur C. Clarke, and this is one of science fiction's earliest and most fully realized depictions of an intelligence emerging from something too small to see. Reads fast, but the ideas hold up as some of the most prescient bio-AI speculation of the 1980s.
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.
