Shareable cards — Thursday, July 23, 2026

Thursday, July 23, 2026 — top 10 database + AI stories, one card each. Save the image, copy its caption, post.

Phone: tap Share image, pick Save Image in the sheet (LinkedIn's app won't take images from Safari directly), then tap the button again — it opens LinkedIn: start a post, add the photo on top, paste the caption. Desktop: Open → save the image, attach it in the composer, paste the caption.

Hand-drawn sketchnote — PostgreSQL from_collapse_limit controls subquery flattening

Open image

PostgreSQL from_collapse_limit controls subquery flattening

The day's edition as one sketchnote — full stories + sources:
https://offnetchat.com/news/2026-07-23.html?s=1f887ad46040417c#item-1f887ad46040417c

#TechNews #AI

AWS Timestream for InfluxDB 3 Enterprise adds multi-node scaling up to 15 nodes

Open image

AWS Timestream for InfluxDB 3 Enterprise adds multi-node scaling up to 15 nodes

▸ Clusters now support max 15 nodes with specialized roles for ingestion, query, and compaction.
▸ Separating roles allows precise resource allocation for mixed workloads.

Summary + source: https://offnetchat.com/news/2026-07-23/s/0f144d31cc7de533.html

#Databases #TechNews

MySQL Galera Cluster hits EOL Sept 2026; Percona XtraDB Cluster is the migration path

Open image

MySQL Galera Cluster hits EOL Sept 2026; Percona XtraDB Cluster is the migration path

▸ MySQL Galera Cluster stops receiving updates after Sept 30, 2026.
▸ New clustering features will only appear in MariaDB Galera Cluster.

Summary + source: https://offnetchat.com/news/2026-07-23/s/4e400c9462d387d3.html

#Databases #TechNews

Startup Postgres Survival Guide: Practical Ops for Scale

Open image

Startup Postgres Survival Guide: Practical Ops for Scale

▸ Prioritize connection pooling to manage resource contention during traffic spikes.
▸ Implement robust monitoring for slow queries and lock contention early.

Summary + source: https://offnetchat.com/news/2026-07-23/s/7e41e843d5a5b295.html

#Databases #TechNews

PostgreSQL from_collapse_limit controls subquery flattening

Open image

PostgreSQL from_collapse_limit controls subquery flattening

▸ Subquery flattening is enabled by default to improve join optimization opportunities.
▸ The from_collapse_limit GUC caps the number of join inputs for flattening.

Summary + source: https://offnetchat.com/news/2026-07-23/s/1f887ad46040417c.html

#Databases #TechNews

TriAgent: Multi-Agent Committees Cut LLM Costs via Divergence Routing

Open image

TriAgent: Multi-Agent Committees Cut LLM Costs via Divergence Routing

▸ Routes queries to cheaper agents when Semantic Divergence Index shows low disagreement.
▸ Combines VADER, FinBERT, and Qwen2.5/Phi-3 agents for granular sentiment analysis.

Summary + source: https://offnetchat.com/news/2026-07-23/s/8fac87b3d4b94135.html

#AgenticAI #TechNews

Anthropic Enforces Deterministic Containment for Claude Agents

Open image

Anthropic Enforces Deterministic Containment for Claude Agents

▸ Deterministic sandboxing replaces soft permission prompts for agent safety.
▸ Trust boundary failures drove recent architectural changes in Claude.

Summary + source: https://offnetchat.com/news/2026-07-23/s/2597471de6eb0354.html

#AgenticAI #TechNews

InfoQ: Jake Mannix on Building Versioned AI Agents with Intermediate Protocols

Open image

InfoQ: Jake Mannix on Building Versioned AI Agents with Intermediate Protocols

▸ Move beyond brittle agent designs by adding an intermediate protocol layer
▸ Implement versioned, encapsulated virtual tools for better control

Summary + source: https://offnetchat.com/news/2026-07-23/s/8d3a051d110bbd9c.html

#AgenticAI #TechNews

Amazon Connect expands agentic voice with 50+ languages and natural pacing

Open image

Amazon Connect expands agentic voice with 50+ languages and natural pacing

▸ Expanded language support covers 50+ languages including Spanish, French, Japanese, and Korean.
▸ New voice options and pacing features make AI interactions sound more fluid and human-like.

Summary + source: https://offnetchat.com/news/2026-07-23/s/512eabb0cc067896.html

#AgenticAI #TechNews

LLM Fact-Checking with Uncertainty Abstention via Evidence Chain Evaluation

Open image

LLM Fact-Checking with Uncertainty Abstention via Evidence Chain Evaluation

▸ LLMs often force true/false decisions even when evidence is sparse or conflicting, reducing reliability.
▸ ECE framework enables abstention via uncertain verdicts instead of forced binary classification.

Summary + source: https://offnetchat.com/news/2026-07-23/s/4c3e3797c50ce45c.html

#LLM #TechNews

S2T-RLHF introduces hierarchical credit assignment to stabilize preference-based RL training

Open image

S2T-RLHF introduces hierarchical credit assignment to stabilize preference-based RL training

▸ Standard RLHF struggles with ambiguous credit assignment when propagating sequence rewards to tokens.
▸ Overly fine-grained reward refinement can destabilize training if preference signals are noisy.

Summary + source: https://offnetchat.com/news/2026-07-23/s/cb178814d1722dda.html

#LLM #TechNews