The memory layer is a retrieval contract, not a storage feature
张洲诚（Zack.ZHANG）
I thought fine-tuning was the hard part
张洲诚（Zack.ZHANG）
I expected the translation model to win
张洲诚（Zack.ZHANG）
Deep research is usually evaluated by reading the report and nodding.
张洲诚（Zack.ZHANG）
Real-time voice is usually sold as a demo
张洲诚（Zack.ZHANG）
Use Bailian's HappyOyster Adventure model to run the full lifecycle:
张洲诚（Zack.ZHANG）
A 46ms Classifier With Zero Output Tokens, and the Calibration Trap in Its Docs
张洲诚（Zack.ZHANG）
A deployment-form divergence, measured across 35 API calls
张洲诚（Zack.ZHANG）
Dot froze on stage, so I re-read a CLI's help text
张洲诚（Zack.ZHANG）
One call makes a full sound scene. Two limits decide whether it ships.
张洲诚（Zack.ZHANG）
Audio model pricing: normalize the meter before you compare the bill
张洲诚（Zack.ZHANG）
The catalog showed zero. The API showed 155: auditing 179 retiring models before October 10
张洲诚（Zack.ZHANG）
A model watched a 76-minute keynote in three minutes: 14 sections, five verified frames, one wrong verdict
张洲诚（Zack.ZHANG）
AI cracked a 370-year-old cipher in 44 minutes: what rebuilding it taught me about verification
张洲诚（Zack.ZHANG）
Two fresh flagships, one prompt: what a head-to-head run taught me about reading model specs
张洲诚（Zack.ZHANG）
The Missing Tool in Your Agent's Hand: Notes From a Hype-Week Field Test
张洲诚（Zack.ZHANG）
I Planted a Fake Policy in My AI Agent's Chat History. It Believed Me.
张洲诚（Zack.ZHANG）
The Same Question Hit Two AI Entrances. One Spilled the Reimbursement Policy. The Other Honestly Didn't Know It Existed.
张洲诚（Zack.ZHANG）