Practices and Perspectives on Data Intelligence
What we've learned in the field — about data, about AI, and about making technology truly serve business.
Interface Wars: Cheap Models, Priceless Data Context
Koa, the AIforce layer and two agent rollouts landed in one day; DeepSeek cut task costs to $0.07. Models are commoditizing — data context is the new moat.
The Next Agent Bottleneck Isn't Models. It's Memory.
OpenAI and Salesforce race on agent orchestration. The real frontier is Iyuno CLOE's persistent ontology graph—memory that compounds, not tokens.
OpenAI's Data Agent: the Real Battleground Is Semantics
OpenAI's Data agent plugs ChatGPT Work into enterprise warehouses, but accuracy collapses to 21% on enterprise schemas. The bottleneck is semantics, not models.
Agentic NL2SQL Just Hit 91.7% on Real Enterprise Schemas
A new benchmark of 900 execution-verified enterprise queries pushes cost-aware agentic NL2SQL to 91.7% accuracy in a single generation — 54.6 points above the runner-up. The real lever was architecture, not a bigger model.
Model Fatigue: A Data Foundation That Outlives Every Model
Four frontier models shipped in one week and CNBC coined it 'model fatigue'. Four design principles for a data foundation that outlives any model release.
Decoding Qwen3.8-Flash-Next: How 51B N-gram Embedding Reshapes the Inference Stack
Beyond the widely-misquoted 58B figure, the 51B N-gram Embedding module is the real story. We unpack its sparse-lookup architecture and address three persistent misconceptions — it is not built-in RAG, it does not retire RAG, and it will not deliver a step-change in general capability. A more measured engineering read on what this release actually changes.
Agents Hit 'Critical': Auditability Unlocks Enterprise AI
OpenAI's Astra is the first model to hit the Critical cybersecurity tier. For enterprises, the adoption gate isn't the score — it's agent auditability.
Data Agents Have Their Own Market Now — 3 Signals from IDC
On September 1, IDC published its first China Data Agent vendor assessment: 18 vendors entered, only four reached the Leaders quadrant, and the firm forecasts that 60% of China's top 500 enterprises will deploy enterprise-grade data agents by 2028. Data agents have officially become their own procurement category. Here are the three signals that matter and the four foundations to lay before going live.
Agents Hit the Runtime Wall: ServiceNow's 9×, Replica Cyber's Isolation, and How OntiCards Got Ahead
Enterprise agents made the same pivot three times in the last week of August: ServiceNow's production-agent customers grew 9× in nine months; Replica Cyber bolted an isolation engine onto high-risk agents; Kyndryl and Google Cloud used a semantic layer plus guardrail agents to lock Swiss Incore Bank's KYC into a governed boundary. Gartner says 40% of agentic projects will be cancelled by end of 2027 — not because agents fail, but because governance is still catching up.