New in Fiitx: top up your workspace with WeChat Pay in one tap — scan on desktop, native pay sheet on mobile. And a new starter-pack layer means new users get working agents on day one, while enterprise workspaces stay exactly as their team configured. #AIAgent
Your agents shouldn't stop just because an API isn't authorized. This release: Fiitx routes invites through UI automation automatically, so bulk creator outreach keeps running on autopilot. And scheduled jobs now appear only for accounts authorized to run them. #AIAgent
Agent harness hardening: connector status no longer conflates 'API not authorized' with 'cannot automate' — UI-automation transports keep working, hallucinated transport overrides fall back to org policy. Tenant cron jobs are now capability-gated: no auth, no visibility, no runs.
'Draft a Thai reply to this customer message' kept routing to the generic chat agent. Root cause: routing ontology matched canonical terms, not real operator phrasing — nor cross-lingually. Added real-utterance anchors incl. Thai; probes show zero cross-domain misroutes. #AIAgent
Someone asked what I'd do if I were 17. I'd learn how to build LLMs from scratch, and then train ones as powerful as I could with whatever hardware I could get access to.
Alibaba is seeking $10.2B from a share sale after quarterly capex reached almost $10B amid its AI buildout.
Says 100% of net proceeds from the proposed HK$80B placement will fund AI capabilities, including infrastructure expansion.
Will make it the largest primary follow-on offering ever by a Hong Kong-listed company if completed.
Alibaba spent $11.9B buying back stock in fiscal 2025 and cut shares outstanding by 5.1%; now it wants $10.2B by issuing new shares for AI. Those decisions don't cancel each other, but they show how quickly capital priorities have changed.
Because infrastructure consumes cash before the resulting capacity is fully monetized, Alibaba is accepting dilution rather than funding the entire next leg from its balance sheet.
Because Alibaba owns Qwen, cloud infrastructure, and proprietary AI chips, it has several ways to earn from AI demand even if model pricing keeps falling.
The Tiangong team responded to the robot’s face-covering victory run: “We didn’t find some young girl to teach it being shy🫣. It came up with the idea all by itself.”
Different sizes, different strides 🤖 Small-sized robots showed a range of running styles, while Tiangong won the large-sized 400m final in 38.15 seconds at the 2nd World Humanoid Robot Games in Beijing. Stay tuned for more highlights from the Games! #Chinarobot#China#WHRG #Beijing
My AirPods Pro 1 started making a loud static noise in the right ear.
I ran a full diagnosis:
— Left ear: perfectly fine
— Right ear: static noise
— Noise cancellation: static
— Transparency mode: static
— Turn both off: silence
— Covered the mic: still static
After all that sophisticated debugging, the diagnosis was:
“Yeah… it’s 5 years old. Just live with it or buy a new one. Don’t bother repairing it.” 😂
Sometimes the most expensive debugging tool is just common sense.
Creator-outreach agent: demo-grade → production-ready. Not a smarter model — a more honest harness: 'already invited' is a legit terminal state, not a tool failure; a local ledger stops re-inviting the same people; every step user-visible. Guardrails beat retries. #AIAgent #Fiitx
Introducing https://t.co/AjkSboF0Af
You don't have to beat a $15k all-time #1 to get seen. Spend a bit today and you can own the top of the board until midnight. Then everyone starts even again tomorrow.
Someone asked what I'd do if I were 17. I'd learn how to build LLMs from scratch, and then train ones as powerful as I could with whatever hardware I could get access to.