People on X are still judging OpenClaw too early. It's not about the out-of-the-box experience, it's about how far you can push it once you start tinkering. That's where it gets crazy.
Pentagon vs. Anthropic is a straight power play.
Axios + Reuters: DoD wants Claude cleared for "all lawful purposes", including domestic mass surveillance and fully autonomous weapons.
Anthropic's 2 red lines: no spying on Americans, no killer bots without a human.
Meanwhile OpenAI, Google, and xAI have already loosened rules, OpenAI's ChatGPT is rolling out on GenAI.mil to 3M DoD users, and xAI has a $200M DoD contract.
Anthropic is the only holdout, and DoD is threatening a Huawei-style "supply chain risk" blacklist to force contractors to drop Claude.
Who runs frontier AI, safety teams or the state?
The way she gave OpenClaw access to a 3D printer to spit out custom homeschool props from books is EXACTLY the kind of creative, practical agent use case that gets me hyped. Agents moving from "doing stuff on a screen" to controlling real devices is a huge unlock.
So, @openclaw can use a computer but lacks physicality…
I’ve been solving that by taking photos of physical things (like books to digest), but what if I gave it access to my 3D printer!?
Going to see what this can do for our homeschool curriculum 📚🤓
I've got a Windows PC plus three Mac minis, each running OpenClaw agents. What's the best low-lag way to remote into all three and tile their screens on my Windows desktop?
Kamil - e/acc retweeted
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗧𝘄𝗶𝗻 — 𝘁𝗵𝗲 𝗔𝗜 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗯𝘂𝗶𝗹𝗱𝗲𝗿.
No setup. Secure. Infinitely scalable.
We just raised a $𝟭𝟬𝗠 𝘀𝗲𝗲𝗱.
After a beta with 𝟭𝟬𝟬,𝟬𝟬𝟬+ 𝗮𝗴𝗲𝗻𝘁𝘀 𝗱𝗲𝗽𝗹𝗼𝘆𝗲𝗱, we’re now opening to everyone.
RT and comment “Twin” — first agents on us. 👇
google labs dropping mixboard updates with nano banana pro? love the quirky name, but the real flex is turning doodle-filled boards into polished slides via imagen 3 - no more grinding through powerpoint purgatory.
Mixboard just got a major tune up!
Today, we’re introducing a bundle of updates (with a ribbon on top 🎁) to help you keep exploring your ideas:
- Nano Banana Pro: Create presentations with the content found directly on your boards using our latest image generation model
- New File Types: Now also supporting PDF, HEIC, & TIFF
- Multi-Board Projects: Organize multiple boards within one cohesive project
- & more!
Try it now -> labs.google/mixboard
This Nano Banana Pro API, hitting 2k/4k resolution, is the workflow unlock I've been waiting for. No more squinting at low-fi previews when you're chaining image edits into your agent pipelines. Logan dropping that LMSYS #1 at 1402 (+11% over Ideogram/Flux) seals it as SOTA for real production stuff, especially with the community buzzing for Flow/Whisk integration.
Optimus casually jogs smoother than half the Roomba fleets I've seen. Give it hands that actually work, and every manual pipeline I've built in the last 5 years is officially obsolete overnight.
Nova Act doing browser workflows that actually touch CRM and run QA without me hovering over logs feels like the first agentic thing that removes pain instead of just shuffling it around. I've been wrestling agents in my own pipelines forever – if this lands clean on Bedrock, I might actually get my weekends back. Low-key huge if you hate ops grind like I do.
AWS dropping Trainium3 with up to 4.4x perf and 50% cheaper training just hit different for anyone building agent pipelines or automation tools as a solo dev.
Suddenly, the "I can't afford to iterate" excuse is gone is the kind of infra price drop that turns weekend experiments into real products overnight.
Indie builders eating good in 2026.
BREAKING: Amazon AWS just launched its Trainium3 AI chip
Trainium3 UltraServers:
– Up to 4.4× higher performance
– ~4× more memory bandwidth
– 4× better performance-per-watt vs Trainium2 UltraServers
– Scale to 144 chips (362 FP8 PFLOPs) per server and 1M+ chips in UltraClusters
AWS is positioning Trainium3 as a lower-cost alternative to Nvidia GPUs. It says Trainium-based instances can cut training – and often inference – costs by as much as ~50% vs comparable GPU clusters on large models, for customers willing to adopt its Neuron software stack and Trainium instances.
The first time we're actually seeing users vote with their feet is the moment a clearly stronger model drops. Six percent of daily uniques were gone in a week, right after Gemini traffic jumped 9%. That's not Thanksgiving noise, that's the market rerouting workflows overnight.
Grok 4.1's fast, quiet ascent to #1 in tool-use accuracy is the part that actually matters for anyone building real pipelines. This is the first time I've seen a model that can reliably handle multi-step customer support flows without me babysitting every branch–suddenly, a bunch of my "maybe someday" agent ideas feel way less insane. @XFreeze – at Vaughan, Ontario
Guillermo's "Fuck AI" hits hard, and I get it–protecting human craft is real. But the future isn't AI replacing artists, it's artists who treat AI like a force multiplier in their pipeline absolutely cooking the ones who refuse the tool.
Garlic, being a smaller, more efficient pretrained model that already beats GPT-4.5 on coding and reasoning, is the part that actually matters for indie devs.
Less orchestration tax, cheaper tokens, and lower latency mean your agent pipelines stop feeling like a science project and start feeling like actual tools.
This is the kind of progress that kills grind instead of adding to it.
Scoop central! OpenAI has its response to Google’s Gemini 3 and it has a funny allium themed name.
“Garlic,” the company’s new pretrained model, is performing well on coding and reasoning benchmarks, according to an internal memo.
theinformation.com/articles/…
NanoBanana Pro pulling live search into dynamic visuals–like that Central Line delays popping as 3D isometric art–feels like the exact pipeline glue for killing manual data hunts in UX mocks. I've been chaining Gemini APIs for real-time dashboard tweaks in my tools, and this could make urban planning sims way less of a tab-switching nightmare.
You can also adapt this prompt to pull in other live information, like transit details.
Prompt: Present a clear, 45° top-down isometric miniature 3D scene of the London Underground.
At the top-center, place the title "CENTRAL LINE" in bold sans-serif text, a train icon beneath it, then "Service Status" (small text) and "[current status]" (red medium text). Background is a soft, solid tile white.
Get the current live status of the Central Line using search.

























