Your AI demo works. What will it take to make it useful in financial workflows? Join the Alibaba Cloud × AMD Financial AI Workshop in Hong Kong: an afternoon on the data foundations for useful enterprise AI—traceable knowledge, durable agent memory, and governed access. Explore PolarDB-powered RAG for finance and RDS Agent Context memory, then build a financial digital-employee prototype in a hands-on session. Free QwenWork credits included. For tech, data & AI teams. Fri, 27 Nov 2026 | 13:30–17:00 HKT | Hong Kong Check-in from 13:30; programme 14:00–16:40. Venue details shared with confirmed attendees. Apply: Places are limited, apply early. #AgenticAI# #AlibabaCloud# #PolarDB# #AgentContext# #FinancialServices# #HongKong#

shared memory without isolation drops agent context when tools write to unpartitioned tables. Cursor clobbers what Claude Code stored at the shared persistence layer. enforcing per-tool scopes in SQLite with sqlite-vec guarantees isolated write lanes.

Long-horizon RL in the multi-agent context seems to converge to the sort of swarm intelligence seen in ants, not humans. Swarm intelligences are inherently harder to control and monitor.

Perfecting the overnight mode on Grok Build to keep the main agent context clean, dropping compaction from 3 times to 1. Prompt: "Overnight mode on: . The role of the orchestrator is to a) make high-level strategy decisions, b) make sure sub-agents deliver useful information and keep going, c) validate sub-agent results (fan out another sub-agent as a validator), and make the next move."

Now in Grok New connectors will be suggested automatically based on what Grok needs to generate more productive results. The agent context will update to include the new tools without a restart. Connections persist across chats.

Practices for embedding AI agents into enterprise systems [Layered Memory -- 4-Tier Memory & Context Broker] 💡 An AI agent that starts from scratch every session is like a new hire who asks the same questions every day. Separate memory into four layers and dynamically assemble "only what's needed right now" -- that's what makes an agent production-ready. 🔥 Problems Solved - Context loss across sessions: context vanishes between sessions and agents, forcing repeated work - Finite context window: stuffing full history into the window degrades both cost and accuracy - Memory bloat: unlimited accumulation increases cost, privacy risk, and context pollution - Lost in the middle: injecting too much context actually reduces answer accuracy 🏗️ Proposed Pattern Separate memory into four tiers: Working (current session, ephemeral), Episodic (past summaries, per-user), Semantic (RAG-indexed knowledge), and Organizational (people, teams, relationships as a knowledge graph). Assign each tier its own storage, TTL, and ACL. A Context Broker retrieves from relevant tiers based on user intent, then prioritizes, summarizes, and compresses within a token budget -- assembling context from only the most relevant information. ✅ Selection Criteria - When to use: agents providing continuous support across sessions and agent boundaries - When NOT: one-shot stateless tasks; use cases where a small, fixed context is sufficient ⚠️ Pitfalls - Episodic memory bloat: store summaries (not raw logs) and control with importance scores, TTL, and time-decay - Context broker quality: poor reranking lets irrelevant information slip through, degrading accuracy - Cross-layer ACL consistency: when layers have different ACLs, aggregation must reduce to the strictest permission 🛠️ Implementation Approach 1. Deploy a vector DB (Pinecone / Weaviate / pgvector) for the semantic memory tier and Neo4j for the organizational knowledge graph tier, configuring storage, TTL, and ACLs per tier 2. Build a summarization pipeline for episodic memory -- store summaries instead of raw logs and implement automatic forgetting via importance scores and time-decay 3. Implement the Context Broker with reranking (Cohere Rerank / cross-encoder) to dynamically assemble the most relevant information within a token budget (target ~8,000 tokens) based on user intent 4. Leverage memory management frameworks (Mem0 / Zep) for cross-session and cross-agent context persistence 5. Tag each memory tier with ACL metadata and integrate with the Context Firewall (P10) to reduce permissions to the strictest level at aggregation time #AIAgents# #EnterpriseArchitecture#

Impressive work from Alibaba. (bookmark it) If you build long-running agents and keep rewriting your memory schema, take a look at this approach. It basically treats agent context management as a programming task. Here is how it works: It backs each agent session with an append-only event log and a sandboxed, persistent Python kernel. Tool outputs, retrieved history, and derived state bind to typed variables across model calls instead of being serialized into the prompt every turn. Model-written code searches and transforms that state, and only explicitly printed projections enter the working view. The event log keeps lossless ground truth, so nothing has to be committed to a compressed form before you know what will matter later. When the working view nears its budget, stale spans are evicted but stay recoverable. An eviction index keeps compact landmarks tied to exact event-log addresses, so the agent navigates straight back to a region instead of searching the whole log. Results: with Qwen3.8-Max, 94.8% on LongMemEval_S, 73.1% on BEAM_10M (5.1 points over the best published memory system), and 86.7% on LOCA_256K. Treating context management as a programming task means it inherits every future improvement in model coding ability. Paper: Track more trending AI papers in our academy:

Credential leaks continue to happen, even though the controls for human identity have been available for years. Matt Moore, co-founder and CTO of @chainguard_dev, joins @NancyzWang and Dev Tagare to discuss how security changes when credentials belong to people, machines, and AI agents moving through the SDLC. The controls differ, but the standard is the same; credentials should be protected and scoped, and every use should be attributable. ✅ People need MFA, secure credential storage, and SSO where it fits ✅ AI agents need short-lived, least-privilege access ✅ Keep raw secrets out of agent context where possible, and broker access at runtime ✅ Trace actions to the agent and its authorizing human or service The gap sits between what teams know works and what's actually deployed. 🎧Listen to the full conversation here: