   TLDR: We built a free web app that helps kids practice navigating comment threads safely before doing it in public. It works like training wheels for social media, giving guided feedback on conflict, privacy, boundaries, and better responses. Next, we are adding classroom multiplayer so students can role-play online situations together and reflect on what they learn One of our passion projects is our Comment Thread / Social Media Simulator - a supervised practice space for first-time internet users. Web-app Link here WHY MAKE THIS? Too many kids are expected to navigate comment threads, exclusion, pressure, and online cruelty without ever having practiced what to do. In our Social Media Simulator web-app, a child reads a realistic thread, decides how to respond, and then gets guided coaching on what was safe, what escalated the situation, and what a stronger response could have been. We think the first lesson in online conflict should happen in rehearsal, not in public. That feels especially urgent right now. Over the last year, regulators have pushed harder on child safety online: the European Commission published formal guidance on protecting minors under the Digital Services Act, then later said TikTok’s addictive design was preliminarily in breach of that law. In the UK, regulators also pressed Meta, TikTok, Snap, YouTube, and Roblox to do more on age checks and child protection. A recent U.S. jury verdict against Meta and Google likewise focused on design harms to young users, not just content alone. The problem is not only what kids see online. It is how online situations pull them in, speed them up, and teach the wrong habits. The U.S. Surgeon General’s advisory noted that nearly 40% of children ages 8 to 12 use social media, even as robust independent safety analyses are still limited. Adults are tired of improvising after the screenshot, after the pile-on, after the feelings are already hurt. We were also inspired by the Center for Humane Technology and the public conversation sparked by The Social Dilemma and The A.I. Dilemma (- big fans of Tristan and Aza!). Their work helped make one thing plain: digital systems are often designed around attention, reaction, and engagement, not healthy development. Finally, this topic just feels important to us at a gut level. Our business is understanding the boundary between technology and ourselves and finding a healthy midpoint. We acted on that intention when designing the right shape and size for the Paletta, i.e. a Paletta is the ideal adapter from iPad to human in terms of comfort, security, wrist-and-neck health, versatility. We believe we ought to give it a try, at least, throw some ideas out and see who calls. WHAT DOES THE APP DO? Our approach is to give children - or anyone really - a safer first step into online life. Our app is training wheels for comment threads. Kids can practice how to pause, read a thread, avoid rewarding cruelty, protect privacy, set boundaries, and know when the strongest move is not to reply at all. They can even write in custom scenarios which are anonymized and turned interactive. In our next update is classroom multiplayer mode, where students can take on different roles in the same thread, including the target, the bystander, the defender, the escalator, or even the bully, then come together afterward to discuss what they learned. It should also help classrooms talk honestly about pressure, responsibility, empathy, and the real social dynamics behind online harm. IMMEDIATE NEXT STEP: Share links and get user feedback. Web-app link here. Stay balanced! Stay tuned! -Team Paletta  We are not preparing young adults for social media chat threads.  Find your feelings with a little guidance. We have been working on Emotion Wheel + DBT AI, a small interactive app designed to help people name what they are feeling with more clarity. Try it here.
The intention is to create a gentle tool that helps someone move from a vague sense of “something feels off” toward more precise emotional language, reflection, and a little more self-understanding. The app includes a large interactive emotion wheel - fun for fidgeters and an homage to the physical version - tap-to-select feeling words, and a DBT-informed AI chat that helps users reflect without pretending to diagnose or replace real support, It also supports custom wheel upload and export, and a semantic discovery view for exploring the emotional map. It is meant to be practical, lightweight, and easy to play with. We released it for free so people can try it, see if it helps, and tell us what feels useful or missing. If it helps you or makes you curious, share it with your friends. -Team Paletta  PCA visualization of 56 political word pairs encoded with all-MiniLM-L6-v2 (384-dim → 2-dim). Red/blue markers denote opposing semantic poles. PC1 (5.7%) captures ideological orientation; PC2 (4.3%) separates affective vs. structural terms. Low cumulative variance (10%) indicates high-dimensional semantic structure—most political language relationships exist beyond these two principal components. SUMMARY
This PCA visualization reveals how political concepts naturally cluster in semantic space. Red and blue markers aren't arbitrary—they trace the dimensional boundaries where ideology meets emotion, where "faith" neighbors "trust" and "despair" sits near "fragile." Notice how "optimism" and "pessimism" anchor opposite poles, while terms like "altruism," "secularism," and "ideology" form their own constellation. The clustering suggests these aren't just words—they're cognitive landmarks in how we structure political thought. The low variance (5.7% and 4.3%) reminds us: political language is high-dimensional. What we see here is just the shadow of something far more complex. NEXT: We will use these words to define new dimensions for other visualizations. Stay tuned!  Two agents (gemma-3-4b @ Temp=1) play 15 games of Mind Meld, with the same initial pair of words, and mostly converge after 3 rounds. We have run 1000s of experiments observing and analyzing how small LLMs (1B to 14B) interact with each other.
One of our favorite experimental architectures is Mind Meld, a fun human improv game where 2 agents work to converge on the same single word, depending only on the previous pair of words. We track the similarity of their words at each round of Mind Meld - creating "similarity trajectories" . We also track *when* convergence happens. Sometimes AI pairs crash through chaos before suddenly snapping into alignment (those stars). Others never quite agree. The question isn't just if they converge—it's how, and what that means. We love it :) WHY IS THIS IMPORTANT? Could multi-agent debates unlock new optimization? Inform decision-making? Spark creative solutions? Self-correct without humans in the loop? We're exploring the edges where cooperation becomes computation, and the data hints at something deeper. We're still figuring out what. Stay tuned for more results and other games of emergence. 
Screen WaitA simple wooden cover that turns a reflex into a choice. Place it over your phone to create a pause — cover the screen, breathe, decide. We keep one at every desk and table. In meetings or at dinner, sliding the cover on helps everyone stay present. Each piece is hand-finished in maple or walnut, with optional engraving for teams or gifts. Screen Wait reflects Paletta’s aim: a healthier rhythm with technology. It’s a calm, human-scale reminder that not every tap needs to happen now. #palettalabs #palettalife #digitaldetox #mindfultech Not everyone wants to code in Python, but almost everyone at Paletta wants to explore agents. Current agentic frameworks like LangChain were too heavy for what we needed. Paletta is still a small company, and needed something lighter, and more ...fun.
WHAT WE DID So, we wrote our own 1) a React IDE and 2) a scripting language to use inside the IDE, optimized for building agentic workflows within Paletta's operations and for exploring emergence. DESCRIPTION Agent IDE (name TBD) is a scripting application with built-in co-programmer optimized for writing and executing agentic workflows. Inside of it, we have access to agent and workflow libraries. We can also write custom agentic-workflow from scratch using our AIDE (name TBD) scripting language. This domain-specific language (DSL) allows everyone at Paletta to write and test agentic workflows with an intuitive language abstraction optimized for our team:
HOW DOES THIS IMPACT PALETTA? Workflows that take care of low-value tasks in our company allow us to focus on the better designs, reliable testing and offer consistent and relevant communication to visitors and customers. Following Paletta's mission, Paletta Labs is also concerned with technologies like mobile devices and AI in society, and wants to stay at the fore-front of practical applications of algorithmic intelligence. RESULTS? Stay tuned for more on Agent IDE/AIDE. 
overlays: Color images of physical world. main: image generations guided by real-world image features.
In Paletta LabsAt Paletta Labs, we like to see what happens when simple materials meet new tools. For this experiment, we set up a small area on the workbench, placed a few grains of rice, and let an image model interpret what it saw. In seconds, the pattern turned into something that looked like a city. It’s a small reminder that a basic gesture can spark a larger idea, especially when our tools are designed to listen. How We Use ItThis setup is part of our internal workflow — a real-time, multi-modal tool we use for brainstorming. It connects cameras, objects, and AI models into one feedback loop. We move materials around, the system generates visual ideas, and we respond to what it creates. It’s less about automation and more about conversation - a way to think visually and prototype quickly across both the physical and digital worlds. Why It MattersPaletta’s philosophy is to build a healthy relationship with technology - one where tools extend our awareness instead of overwhelming it. Projects like Architecture from Rice remind us that creative systems don’t need to be complex to be powerful. A little rice, a lens, and a responsive model are enough to start reimagining how we see and design. #palettalabs #architecture #aiworkflow #diffusionmodels #genai TLDR: Organizing your team's Palettas on a wall to save space while charging + Customize the angle of access to match the vertical location in your team's space.
FEATURES: Mount and display multiple Palettas (photo shows 10-capacity). Angled for ergonomic access. Easy installation with French-cleat design. Guides for cable management. AVAILABILITY: Custom request by email, [email protected]. |
RSS Feed
