|
|
Welcome, humans. |
Official back to work week after the holidays is coming in rough this Tuesday. We were low-key enjoying not having a flood of AI news to sift through every day over the last two weeks (just kidding, it somehow still flooded AI news over the holiday break, just like… not as hard??), and then CES hit us like a ton of bricks yesterday. |
The vibe this year? As best we can tell via armchair quarterbacking the show from afar, AI is going everywhere: your neck, your glasses, your pets, your cat's food bowl, even your haircut, heck, even a piece of wood! |
It's also apparently going straight for your jugular: |
 | TBH, we have no idea if dropkick bot is at CES, but if it is… check your necks ppl! |
|
Whether we needed AI in all these places is… debatable. For example, do we really need robots that can drop kick you in the face harder than Chuck Norris? 6+ Terminator movies and counting say otherwise. But hey, at least that cat health monitor seems genuinely helpful (and on brand for us). |
Here's what happened in AI today: |
NVIDIA opened CES 2026 with its own flood of new releases. OpenAI said 40M+ people globally use ChatGPT daily for health info. Alibaba's PANDA AI detected ~24 pancreatic cancer cases in 180K CT scans. ex-Apple Face ID engineers raised ~$107M to develop robot vision technology.
|
|
We need your help. We're finalizing our roadmap for The Neuron in 2026, and we want YOU to weigh in on what you want more of in 2026. So tell us: What do you actually want from us this year? Do you like livestreams, or deep-dive articles? Do you want hands-on workshops or more industry-specific content? In-person community meetups? Or courses? Or all of the above? Your answers directly influence what we build this year. Take the survey here. P.S. We'll also be drawing a winner from the contest we ran promoting this survey, to be announced this week or next, so stay tuned to see if you won some quality time with ya boys! |
|
|
Don't forget: Check out our podcast, The Neuron: AI Explained on Spotify, Apple Podcasts, and YouTube — new episodes air every week on Tuesdays after 2pm PST! |
|
NVIDIA Built a Self-Driving Car That Can Explain Why It's Braking |
|
NEWS BRIEF: Everything we've clocked from CES 2026 (so far)… |
At CES 2026, Jensen Huang took the stage and dropped a line that might define the next decade: "The ChatGPT moment for physical AI is here." Bold claim, boss. |
The star of NVIDIA's keynote was Alpamayo, what Jensen calls "the world's first thinking, reasoning autonomous vehicle AI." |
Here's the problem it solves: |
Traditional self-driving cars separate seeing from deciding. Works great 99% of the time, but that remaining 1% (weird edge cases, unexpected scenarios) is where accidents happen. Alpamayo does something different; it reasons about what it sees. It can explain why it's making a decision, describe the scenario, and think through novel situations step-by-step, like how you'd talk through a tricky merge with a nervous passenger. Huang explains: the system doesn't just output steering commands, but tells you what action it's taking, why it chose that action, and the trajectory it plans to follow.
|
So what will be the first car with Alpamayo? The Mercedes-Benz CLA (conveniently just rated "the world's safest car" by the NCAAP), shipping Q1 2026 in Europe. It runs two AI stacks simultaneously: Alpamayo for reasoning, plus a classical backup system. Redundancy for safety. |
What else NVIDIA announced: |
Vera Rubin: Next-gen AI supercomputers with 1,152 GPUs per pod, cooled with 45°C water (no chillers needed; v chill indeed). Nemotron Speech: Voice recognition that's 10x faster than competitors—Bosch is already using it for in-car assistants. Isaac GR00T N1.6: Foundation models for humanoid robots, now with full-body reasoning. Cosmos: Physical AI models topping reasoning leaderboards—Uber, Salesforce, and Hitachi are building on them.
|
Why this matters: If you've ever been nervous about trusting a self-driving car, this (potentially) changes the game. These cars will both drive you and explain why they're driving that way. When something goes wrong, there's an actual reasoning trace to examine. Not a black box. |
Our take: This will be an interesting spoke thrust into the wheel(s) of the Tesla vs Waymo self-driving race, which recently took its own unique turn. While Waymos are definitively safe enough to roll out across many major cities, a recent blackout in SF proved they're not so reliable without a city's infrastructure to support them; meanwhile, Dave Moss just drove a non-robotaxi Tesla 100% coast to coast across the US in full-self driving mode with 0 disengagements (including for parking). |
So we might be looking at a city car vs country car debate here… if Apalmayo can operate without the infrastructure Waymo needs, we could have another cross country competitor ready to revv its engines (and tell you precisely why it revved said engine)… |
|
FROM OUR PARTNERS |
How Elastic helps Sitecore simplify content management |
|
Imagine reaching into an ocean of content and pulling out exactly what you need in the blink of an eye. Elastic helps Sitecore customers do just that. |
Sitecore gives its clients the power to manage millions of marketing assets, from images to videos. |
By incorporating Elasticsearch into its content hub, Sitecore makes content discovery a breeze, even at massive scale. And with Elastic Security's role-based permissions, assets are always protected against theft and incorrect use. |
That kind of search power allows Sitecore clients to accelerate campaigns — and makes Sitecore a critical part of their tech stack. |
Check out the full story→ |
|
Prompt Tip of the Day |
Stop negotiating your boundaries in real time: write them once, then reuse them. |
Think of this one like a Personal "Terms of Service" to turn fuzzy intentions into something you can actually follow when you're tired, busy, or being guilted into "just one more quick thing." It's basically a friendly contract with future-you: clear defaults, fewer awkward moments, and less calendar chaos. |
Drop-in prompt: |
"Ask me questions about my goals and help draft my Personal Terms of Service for 2026. Note my role(s), priorities, dealbreakers, ideal week, and 3 pet peeves. Keep it 1 page, lightly funny, and include 3 copy/paste boundary scripts. Write it like product rules: short sections, bold headers, specific behaviors.
|
|
Treats to Try |
Qwen‑Image‑Layered (yes, again) is worth bookmarking if you make visuals: the layered output makes real edits feel like Photoshop instead of "prompt and pray."—free to try. Amazon launched Alexa.com, bringing its Alexa+ assistant to your browser—free with Prime after early access. Some use-cases: Type "create a week of meals" and it generates options with your dietary preferences then adds every ingredient to your Amazon Fresh cart. Drop recipe links to customize for dietary restrictions Upload documents to extract appointments and add to your calendar. Control your smart home in the same window.
Google Edge Gallery lets you try models that run entirely on your iPhone (via TestFlight), so you can chat, use images, and even ask with audio without sending data to a server once the model is loaded—free to try. Rork just launched version 1.5 with RevenueCat integration (prompt "add a $9.99 paywall" and it handles A/B testing for ya), built-in analytics that track daily/weekly/monthly users without Firebase, an upgraded AI agent powered by Opus 4.5 and Claude Code, an iOS simulator that emulates every native device in your browser, and App Store-ready icon generation—all for building mobile apps from text prompts. Diffs renders code changes with custom annotations and AI-generated UI on any line that sync automatically across split and stacked views—open source. SkillsBench benchmarks whether giving agents specific "Skills" actually improves their performance and how well they use them—open source.
|
|
Around the Horn |
 | We knew it! |
|
OpenAI said 40M+ people globally turn to ChatGPT daily for health information, per an analysis of anonymized interactions, seeing it as an "ally" for navigating health care and insurance, with ~70% of conversations happening outside normal clinic hours. Google plans to integrate Nano Banana and Veo support into Gemini on Google TV, letting users generate AI videos and modify family photos with voice commands, first rolling out on select TCL TVs before broader deployment. A Chinese hospital said Alibaba's PANDA AI tool has analyzed 180K CT scans and detected ~24 pancreatic cancer cases since its deployment in November 2024, including 14 in early stages, with some cases flagged by AI that doctors initially missed. Lyte, founded by ex-Apple Face ID engineers Alexander Shpunt, Arman Hajati, and Yuval Gerson, emerged from stealth and raised ~$107M to develop tech to help robots see better and move more safely in the world. A recent paper found that LLMs are systematically overconfident about their ability to solve tasks, predicting higher success rates than they achieve even as model capability increases; a limiting factor for future autonomous systems in safety-critical settings.
|
|
FROM OUR PARTNERS |
Ideas move fast; typing slows them down. |
|
Wispr Flow turns your speech into clean, final-draft writing across email, Slack, and docs. It matches your tone, handles punctuation and lists, and adapts to how you work on Mac, Windows, and iPhone. No start-stop fixing, no reformatting, just thought-to-text that keeps pace with you. When writing stops being a bottleneck, work flows. |
Give your hands a break ➜ start flowing for free today. |
|
Tuesday Tool Tip: Use SAM 3D to turn any photo into a 3D model in 30 seconds. |
|
Creating 3D models used to require expensive software and hours of manual work. Well, Matt Berman just shared his experience using Meta's new Sam 3D tool to create a model that he could then 3D print. See, Meta's SAM 3D collapsed that entire workflow into a few clicks… and it's completely free. |
Here's how it works: SAM 3D extracts objects from regular photos and converts them into downloadable 3D models. No CAD experience required. To try it yourself: |
Upload any image to Meta's SAM Playground. Product shots, video game characters, buildings—anything. Click the object you want. The AI highlights it automatically. If it misses something (like a glass lid), click "add" and tap what's missing. Use "remove" if it grabbed too much.
Pro tip (from the SAM team's AMA): Use simple one-word prompts like "speaker" or "lamp"; the model works best with single noun phrases. Hit "Generate 3D." Wait a few seconds, then download as PLY or GLB format.
|
BTW: You can also do this locally with ComfyUI and Hunyuan 3D 2.1 locally (try it here). |
What you can do with it: |
3D print it (Meta's team tested prints at ~1 inch scale—works great). Import to Blender for editing and refinement. Drop into game engines or VFX workflows. Use for product prototyping.
|
To create complex scenes (multiple furniture pieces, architecture): Segment each object individually with SAM 3 first, then reconstruct each separately with SAM 3D. Combine them afterward; you'll get way better resolution than processing the whole scene at once. |
Bonus features: |
3D Bodies tool: Detects multiple people in photos and maps their full skeletons—even through occlusions. Perfect for animation rigging. Effects: Add shimmer, change materials, apply visual filters once generated Fine-tunable: Works with just ~200 training samples for specialized use cases
|
Reality check: The backside of the image gets approximated (no camera angle means, ahem, educated guessing). And 3D-Body meshes come out cleaner than 3D-Object meshes, so use post-processing tools if you need production-ready topology for general objects. |
|
|
 | Who wants to tell him?? |
|
|
| That's all for now. | | | What'd you think of today's email? | |
|
|
No comments:
Post a Comment
Keep a civil tongue.