The infrastructure plumbing determines exactly how far an AI pilot can scale. When legacy dependencies and data bottlenecks clog the system, even the most advanced models fail to deliver production value.
SYSTEMS
-
Agent governance directly in LangSmith
By
–
Agent governance is not something that should be added to agentic systems. LangSmith LLM Gateway allows you to apply rules on the same platforms where agents are built, observed, and evaluated.
-
Long video reasoning with DeepSeek sparse attention and multimodal MoE
By
–
"Kwai Keye-VL-2.0 Technical Report" This paper makes long-video reasoning much more feasible by adapting DeepSeek Sparse Attention to a GQA-based multimodal MoE, reaching 256K context with only 3B active parameters. As dense attention makes hour-level context way too expensive,
-
Multi-step Workflows: Persistent State and Parallel Bursting
By
–
You run multi-step workflows where state must persist across tool calls. You need bursting capability (i.e., thousands of parallel environments for RL training or evaluations) that must go from zero to scale in
-
Mythos-class models: 4-8 months to harden IT systems
By
–
Assuming open models continue to lag about 8-12 months behind closed source (at least in coding), the countdown to hardening IT systems against Mythos-class models is now at 4-8 months Having publicly available and relatively safe defensive Mythos-class models today is important
-
Codex internal memory API endpoints for managing notes
By
–
not exposed to the end user (for good measure) but as an in-built tool that codex can call it can do the following: memories/list
memories/read
memories/search
memories/add_ad_hoc_note -
Deep Agents Part 2: Context Management
By
–
Deep Agents deep dive Part 2 | Context Management
— LangChain (@LangChain) 16 juin 2026
A <2 min explanation on one of the most important capabilities in the Deep Agents harness from @SydneyRunkle pic.twitter.com/tnIsx9aiLiDeep Agents deep dive Part 2 | Context Management A
-
Improving AI Systems without External Control and Managed Costs
By
–
We work with companies and governments worldwide to ensure their AI systems operate and remain operational outside any external control, improve with each new model version, and benefit from a cost structure
-
Agents need one system of record with unified governance
By
–
Without the right context, even the most advanced agents fail to give reliable answers. And when context is incomplete, AI will fill the gaps. “Agents need one system of record, with all of your data and unified governance,” says Databricks co-founder and CEO, @alighodsi
. -
NVIDIA Blackwell platform dominates MLPerf Training 6.0 benchmarks
By
–
The NVIDIA Blackwell platform just swept MLPerf Training 6.0, delivering fastest performance and largest scale.
— NVIDIA (@nvidia) 16 juin 2026
Beyond the benchmarks, capabilities like the Reliability, Availability, and Serviceability Engine and NVIDIA Resiliency Extension deliver fewer interruptions and… pic.twitter.com/QH77j4UA8nThe NVIDIA Blackwell platform just swept MLPerf Training 6.0, delivering fastest performance and largest scale. Beyond the benchmarks, capabilities like the Reliability, Availability, and Serviceability Engine and NVIDIA Resiliency Extension deliver fewer interruptions and