Blog
When writing a Prompt or Skill, don't use an "introductory word + colon"
When writing Prompts and Skills, there is a highly counterintuitive phenomenon: many people like to write rules as "a short tag plus a colon plus a detailed description," thinking this makes the key points stand out and the structure clear, but in reality it actually causes a significant drop in model instruction adherence.
They call reposts by influencers organic traffic
It's completely understandable to want to polish your image after making money, but taking the monetization results that come with a built-in reservoir to teach people who don't even have plumbing how to conjure water out of thin air is a bit too pretentious.
No sooner had I finished complaining than Google officially announced Gemini 4 Argon.
I was wrong. If you often walk in the rain, how can your shoes not get wet? I had just complained about the Gemini team's snail's pace when they turned around and came at me with Gemini 4 Argon to slap me in the face—too swift, completely catching me off guard.
Slow-moving Gemini team launches Skills to replace Gems
Today, Google announced the launch of Skills in Gemini to replace the previous Gems. Skills was already available in Spark Beta, but honestly, the web version had all kinds of bugs and the experience was really terrible; only the Mac client offered a somewhat better experience.
Sarcasm +1, OpenAI has yet again released GPT-6.1 Sol
OpenAI claimed yesterday that, for safety reasons, out of a profound sense of responsibility for humanity's future, and to prevent an out-of-control AI from destroying humanity (I added these two parts), it was pausing the release of the latest Astra. These days, if a model doesn't carry safety risks, it's almost embarrassing to call itself a cutting-edge frontier model.
OpenAI: Reopens $200 Subscription, Same Price for Less
In one sentence: the price stays the same, and the quota is halved. That's not the point; the point is Tibo's wordplay here—it's worthy of a high-end player.
This is deeply ironic — Anthropic has released Claude Sonnet 5.5
How long has it been, again, since the major vendors last called on the industry to slow down and repeatedly warned that AI was about to destroy humanity? Less than a month. At the start of the month, it destroys humanity; by the end of the month, humanity keeps moving forward.
"Faraday Cage": Anthropic Partners with NVIDIA to Launch Claude Managed Agents and OpenShell
Anthropic yesterday jointly released an enterprise-grade Agent security solution with NVIDIA, combining Anthropic's managed service Claude Managed Agents with NVIDIA's open-source runtime security system OpenShell.
Claude subscribers can now submit self-developed plugins.
Anthropic has launched the Claude plugin developer portal, where developers with a paid subscription can submit their own plugins directly to the official plugin library and track security scans, review progress, and post-launch usage statistics.