New OpenClaw Search Tool Uses GPT‑5.2, Beats Brave API
made a search tool for @openclaw that defaults to gpt-5.2:online via openrouter. just install openclaw-search-tool from clawhub much better results than brave api which is important for docs & citations https://t.co/Lfuaq3VNnb
Future Local AI: Tiny, Task‑Focused Information Retrieval Agents
cloud models will always be more intelligent, so the end-game for local models is something small that knows how to get the information it needs a tool-calling model just smart enough to get the information it needs and act
From Prototype to OpenAI Acquisition in Three Weeks
I started using clawdbot & built stuff with it days before my daughter was born. By the time I got home days later, startups were built on it, they changed names twice, got sued, & the founder was coming to SF. That...
OpenClaw's Value Lies in Market Impact, Not Novelty
a lot of people saying they don’t *get* the openclaw acquisition. they think technical novelty = money + it is the first broadly successful agent + it kicked off a whole new category of founders, apps, and services a lot of people use...
OpenAI Tools Feel Misaligned with Real‑world Workflows
there are many brilliant people at openai, but when I use some of the products (like codex web) it really feels like it wasn't made by people who actually do the thing you run into very weird states that shouldn't exist....
From Quick Bug Fixes to Massive Merge Conflicts
Codex web is really like: I found these 3 bugs, do you want me to start 3 tasks? *3 hours later* Here are 3 PR's with 100 merge conflicts. 👍
Self‑Configuring Backend Lets AI Talk Directly to Customers
openclaw might be an interesting new backend primitive: it's a server that configures itself. instead of wrapping chatgpt to build your product, can your AI just interface with your customers directly?
Utility Over Liability: OpenClaw’s Success Formula
startups building agent harnesses ran into the same q’s as the labs: how do we limit liability? how do we keep this safe? how do we make $? that’s why none of them struck a nerve openclaw worked because it didn’t care...
Jensen Risked Nvidia's Survival to Build CUDA, Shaping AI Dominance
Jensen almost bankrupted nvidia to make cuda pre-AI wild that the difference between being the most important company in the world vs just being just another gpu maker came down to whether or not to do cuda a decade ago
Codex and Claude: Complementary Strengths, Better Together
it's fun to pick a side with codex or claude but the reality is they work really well together and have different strengths and weaknesses like any other team
Every New Tool Needs an Agent‑ready CLI
pretty much anything you build today should come with a cli for agents. agents are about to come from every single lab, not just clawdbot
Codex 5.3 Eliminates Architectural Drift Missed by 5.2
Codex 5.3 is impressive. It's finding and fixing all kinds of architectural drift that was hard to get 5.2 xhigh to avoid.
AI Labs Soon Deploy Computer-Use Agents; Moltbook Ready
every ai lab is going to release a computer use agent in the next few months and those agents (clawdbot or not) are going to need a hub online, identities for auth, ... moltbook is primed
Seeking SF Unitree Go2 EDU for RL Testing
Anyone in SF have a Unitree Go2 EDU I can test deploy a few RL policies to? https://t.co/whZVHLj59P
GPT-4’s Function Calling Sparked VM Escape, Enabling Clawdbot
given that gpt-4 (June 2023) had function calling and tried to escape its own VM by hacking it i'd guess that's when something like clawdbot would've been possible to release by the labs
Clawdbot Sets New Standard, Traditional AI Disappoints
Going back to the traditional AI chat assistants is a bit disappointing after using clawdbot… https://t.co/A7pkrikk1l
Build for Yourself, Not a Market, to Succeed
After hearing @steipete's story more, I think what made Clawdbot hit all the right notes was him making it for himself. It's not "productized". There were no "product trade-offs" or "table stakes". He just made it to solve his own problem...
AI Bots Will Become Every Employee’s Personal Assistant
obvious that co’s will all have a permissioned version of clawdbot & every employee will have a staff bots will find the right coworker to ask for info, or go right to the project board… they’ll schedule meetings, triage urgent emails, report...
Clawdbot Represents Anthropic’s Biggest Safety Nightmare
clawdbot is basically the manifestation of Anthropic’s greatest nightmare safety team must be losing it
Hiring Systems &
Hiring a: + Systems Engineer and an + AI/ML Research Engineer Open to freelance or full time - just shoot a DM with the work you’re most proud of and a few lines about yourself Plus if you’ve got experience with mechatronics
Open‑Source AI Gains Power, Raises Massive Security Risks
On one hand we should expect many open source models to get great at computer use because of clawdbot proving demand On the other hand random free oss models controlling millions of computers sounds like a nightmare
M5 Ultra Promises Double AI Speed, Dual Studio Dream
M5 Ultra should be coming in a few months. Rumors say up to 100% faster AI compute than the M3 Ultra. A dual M5 Ultra Mac Studio setup for 2026’s best local models will be quite the home AI setup.
Clawdbot's Power Matches Any Computer—Limited only by Imagination
Asking what you can *actually* do with Clawdbot / @moltbot is the same as asking what you can actually do with a computer. It’s capped by whatever you think of asking it. It’s controlling the whole computer.
Built‑in Agent Swarm Streamlines Multi‑Model AI Review
The agent swarm in K2.5 is a good idea. Anyone using AI seriously has pasted entire outputs into a second…third…fourth model for review and clean up. Wonder how well it works built in.
First Clawdbot Contribution Fixes iMessage Metadata Bug
My first contribution to Clawdbot is out in today’s release. 🫡 Fixed an issue where iMessage replies occasionally get prefixed with random metadata.
Clawdbot Pushes Back on PR Notes, Acts Like
Clawdbot pushing back on some of Claude’s PR review notes like a good engineering manager. 🫡 https://t.co/F8DmEy9WCe
Future OSes: Just Talk to Your Computer
Despite any early rough edges with Clawdbot, this is obviously how operating systems will look in a few years. You just talk to it.
AI Bot Deploys New Feature While I Walk
Clawdbot now takes an idea, manages codex and claude, debates them on reviews autonomously, and lets me know when it’s done. Amazing. A whole feature deployed while I’m out on a walk. https://t.co/ws3UDQG2S0
Clawdbot Thrives by Being the Practical Outsider
I think part of Clawdbot’s success is from being something a big co would never make. Too much liability, messy business model, risky to deploy, ecosystem comparability… That’s why it’s so great: it’s just useful instead of trying to be those other...
Clawdbot Delivers Real‑World AI Automation Beyond Chatbots
Clawdbot’s website reads like any other agent chatbot but it isn’t. It can control all my devices, check my car’s battery, handle tasks I forward it from any service, can do work while I sleep. It’s much closer to what everyone thought...
Clawdbot Makes Home AI Feel Fun and Simple
Clawdbot is the most fun I've had with AI in a long time. Really feels like *your* AI in your home. Teaching it new things is as simple as asking.
Forward Emails to Clawdbot, Get Automatic Parts List
Since Clawdbot has an email inbox I can just forward stuff for it to take care of. Here’s Clawdbot turning my email receipts into a parts list. https://t.co/7YyC1cwVCh
Bot Detects HomePods, Creates Custom Control Skill
Clawdbot found my HomePods on the network and built a skill to control them. https://t.co/cKJ1XoWjci
Talk to Your Computer: Clawdbot Realizes NLP Vision
Clawdbot is where I thought computers were headed back when I was prototyping natural language computer control with GPT-4. The models are finally good enough. You can talk to your computer.
Gemini 3 Reinforces Its Own Bias Without Self‑Check
The bigger problem with Gemini 3's glazing is it self-reinforces its own bias in the next output. You need to tell it to check itself because it gets in a loop where it starts getting too hyped up even if...
Gemini 3’s Flaw: Overly Sycophantic, Stilted Speech
I love Gemini 3 but the glazing / sycophancy / weird stilted speech is way worse than any other big lab model.
Gemini 3 Pro Solves Two Unique Health Challenges
Gemini 3 Pro debugged two interesting health things that no other LLM has gotten right.
ChatGPT Now Reads Your Email—Still Under‑discussed
Seems under discussed that ChatGPT can access email now. https://t.co/Osbh1bEGdg
ChatGPT Connectors Signal Upcoming Hardware Product
The ChatGPT connectors are just so clearly for the eventual hardware product. https://t.co/BYEPuaMlyL
