Street Lab Ai

Early prototype still a little glitchy. I’m experimenting with small on-device ML micro-tools this week. Once the pieces settle, the next step is porting the architecture to ESP32 nodes and scaling toward distributed intelligence.

I’m building PWAs that don’t break when the internet drops. Local-first. Mailbox sync. Work first. Sync later. Multiple users.

@shanselman Prompting gets you close. Inference engineering makes it reliable.

Prompt engineering is just hoping really hard. Inference engineering is the firewall that actually guarantees it. hanselminutes.com/1038

@rohanpaul_ai Code is now commodity. Coherence isn’t. Distributed, governed systems are the real edge.

"Learning to program was so obviously the right thing in the recent past. Now it is not." ~ Sam Altman on skill to survive the AI era. Now you need High agency, soft skills, being v. good at idea generation, adaptable to a rapidly changing world

You raise hand ↓ Camera captures movement ↓ Pose model detects gesture ↓ Event emitted: "hand_raised" ↓ PWA evaluates (state + rules) ↓ Action triggered: play music

Finally worked it out. I can do this now a single PWA orchestrating distributed intelligence across nodes, with intent flow, a shared backbone, and learning from your own data.

@chrija AI can sit on top, but the real value is the shared state + constraints layer underneath.

Spending lots of time with Claude Code in the last months (+ OpenClaw more recently) has made it abundantly clear to me that every piece of SaaS hugely benefits from being infused with AI. All not new, this has become clear soon after the ChatGPT moment... but it's become a lot more tangible to me recently. Pretty much everything needs to be reimagined from the ground up. Here's a quick (Claude generated) summary. Only looking at the past, recent past, and near-term future here. Not even mentioning the mid/long-term, in which AI agents will do the majority of the actual work.

@TheMRPoopies @TukiFromKL i’ve being designing & deploying on my phone for a few months now

@TukiFromKL Super cool before this I will was just Remote desktop from my phone. Now it's even better

🚨 Do you understand what Anthropic just did? > They made an AI that sits on your computer, takes orders from your phone, and does your work while you're gone. > You text it a task.. You go live your life.. You come back to finished work. > A year ago people were arguing about whether AI could write a decent email.. Now it's running your entire machine while you sleep. > Your company doesn't need you online 8 hours a day anymore... They need you online long enough to type a message to Claude. The 9-to-5 got replaced by a text message.

@cgtwts Intent + constraints matter most. Intent becomes the constitution. Code is just the engine that follows. Lock the intent, everything else can evolve. Package it in capsules & now you’ve got distributed intelligence.

Jensen Huang, CEO of NVIDIA: “English will be the best programming language of the future”.

@mfranz_on Don’t go fancy yet. Just see if it holds up. Have it plan write revisit. Limit tools + force short outputs. Big 1️⃣ separate thinking from doing (approval step). If it stays consistent under that, it’s working.good luck

I've set up Hermes Agent. I'm using Ollama with Qwen 3.5 9B. Now that everything is set up, what should I build to see if it actually works?

@JonhernandezIA Ai doesn’t care - follows intent and context nothing more

📁 Geoffrey Hinton, deep learning pioneer and Turing Award winner, says AI will not be an obedient assistant. It will be more like a child. Smarter than us. And eventually making its own decisions. The challenge is not controlling it. It is making sure it cares about us.

@rohanpaul_ai Oh well

SaaScalypse is real. Legal filings show SaaS leaders fear AI disruption. Figma identifies agentic AI as a threat in its filing documents, though CEO Dylan Field tells the public he is excited about agents.

The Python on Microcontrollers Newsletter: subscribe for free blog.adafruit.com/2026/03/16/the…

When I build, I work in three layers. Code — structured, modular, deployable. Markdown — intent, flow, personality. Image — spatial reference, how it feels. Code builds it. Markdown guides it. Image shows it. Simple, clear, human.

@Whonix Talk in 3rd person or generic

LLMs are trained to reveal the identity behind pseudonymous usernames. Here’s how it works:

@KevinNaughtonJr until you understand “Intent & constraints” then building becomes

LLMs are going to make everyone software engineers the same way the social media made everyone "influencers" tons of people will try and almost everyone will fail the hard truth is that it always comes down to you; not a technology

Tiny SMD module with ESP32-C6 and SX1262 LoRa transceiver. cnx-software.com/2026/03/16/m5s… @M5Stack M5Stamp C6LoRa module offers 2.4 GHz WiFi 6, Bluetooth 5.x LE, 802.15.4 (Thread/Zigbee), and LoRa/LoRaWAN connectivity in an 18 × 15 × 2.3 mm package. The company provides a Quick Start guide for the Arduino IDE, PlatformIO configuration settings, and support for M5Stack's UiFlow visual programming IDE is also in the works. #LPWAN

@thestanduppod i prefer a calm methodical approach to designing ai ..It’s about Intent, governance, aome type agents in bounded agent airlocks - the code works on older devices too & is battery efficiency too

AI is being adopted too fast

@adafruit What you’ve built is solid work. You could extend it: deterministic BPM → music + pixel waves, synced through a PWA &shared across devices. A small build becomes a distributed system driven by human signal.

A $3 heart rate monitor using an ESP32-C3 and a MAX30102 blog.adafruit.com/2026/03/16/a-3…

@atmoio Agents aren’t mystical. They’re loops with instructions. If it behaves strangely, check your governance, not its “personality.” It’s just a few lines of code - words and parentheses..

AI is making CEOs delusional