Searxly is actively working on redesigning search results to be minimalistic while retaining essential information. To achieve a more premium look, we’ve transitioned from round mode to flat mode. Here’s the first version of the search results (this is not the final design; we’ll keep you updated on our progress).
We anticipate deploying a commit on our open-source repository tonight containing all the work we’ve completed today.
We’ve been quietly building Local AI Chat into Searxly, and we're introducing it to you while still working on it actively.
It’s a private, on-device chat (powered by Apple Intelligence) that lives right inside the browser. You can open it from the header, from the home page, or after a search.
The first real custom tool we made is this: web search, but routed exclusively through your private SearXNG instance. Nothing goes to any cloud.
How the controls work:
- There’s a master “on-device AI” switch and a separate “AI tool calling” toggle in Settings → Local AI.
- When tool calling is on, the model can proactively decide to search when it needs fresh info.
- Even when the toggle is off, if you explicitly say something like “look this up on the web” or “search for X”, it will still detect the intent and show a clear permission prompt before using the tool.
- You always see exactly what it wants to search and can approve or deny. Every tool use is also logged in the AI activity section.
We’re actively expanding this, more tools, better natural outputs, tighter integration with your local data and RAG options, and more granular controls. The goal is to give the AI real capabilities while keeping every decision and every piece of data under your control.
We’re still early and will soon open source this in the repo.
Below is a demo of how it currently works.
Manila is live. 🇵🇭
The first next-gen Gitlawb node:
https://t.co/42wk0crn9v - now serving the network from AWS:
▸ Terraform-provisioned, one apply from zero to node
▸ RDS Postgres + encrypted EBS + daily snapshots
▸ Auto-renewed TLS, CloudWatch alerting
▸ Images built & shipped from our own ECR
Made possible by an @AWSstartups credit grant.
The network grows. 🌱
Gm fam @officialbunnyos
After going through the latest BunnyOS updates, I think the project is moving in a very strong direction compared to most AI agents on Base.
✅ GMGN integration into the Research tab brings:
🔹 Real-time volume tracking
🔹 Smart money flow insights
🔹 Holder distribution analysis
🔹 Liquidity monitoring
🔹 Risk scoring
✅ BunnyOS is more than just an AI chatbot.
The stack includes:
🔹 Natural language AI chat
🔹 Autonomous scanners
🔹 Markdown-based memory system
🔹 Tabbed terminal UI
🔹 Built-in token research and analysis
✅ Fully open-source:
🔹 Self-hostable
🔹 No private key storage
🔹 Encrypted data architecture
🔹 AGPL-3.0 licensed
✅ v0.3.4 focuses on strengthening the foundation with:
🔹 GMGN-powered research tools
🔹 Improved user experience
🔹 Language fixes across devices
🔹 Preparation for v0.4 UI and trading endpoints
📌 What stands out most is that BunnyOS is positioning itself as a true AI-native agent, not just another tool aggregator or wallet-monitoring bot.
If the team maintains its current pace of execution, BunnyOS could become one of the most interesting AI agent projects in the Base ecosystem. 🐰
This is an example of an Underly bridge for our deployed app chain. This is live https://t.co/pet68eBUC7
Underly bridges are built to let users test the viability of their deployed chains immediately after launch.
Once a chain goes live, teams need a simple way to move funds, test transactions, and confirm that the deployment works beyond just having an RPC or explorer.
We have deployed a mini version of the bridge for early testing, and the UI will be upgraded soon.
The goal is simple: every Underly chain should be usable right after deployment, not just technically live.
New launch, new bag. 👀
Aped into $UND on Base by @underlyio
Underly is building the infrastructure that lets projects launch their own chains without the headache of building everything from scratch.
• Ethereum L2s
• Optimism L3s
• Base L3s
All from a single dashboard.
Think Shopify, but for blockchains.
As trillions of dollars eventually move on-chain, the winners won't just be the assets being tokenized, it'll be the infrastructure making that transition possible.
That's the thesis.
@underlyio
CA: 0xc5267C0DFbB3f1b28F8C0FC59233D28cb1D1Db07