Hugging Face explores benchmarking open models for agentic capability against your own tooling.
Latest signals
Every signal tracked from the world's AI conferences — 4,665 talks, quotes & announcements, newest first.
GLM-5.2 is now the leading open weights LLM on Artificial Analysis's Intelligence Index, released under MIT license.
“GLM-5.2 is the leading open weights model on the Intelligence Index v4.1.”
Apple's Core AI lets developers run cutting-edge AI models on-device for private, offline, zero-cost inference.
“Inference happens on device so data stays private, AI features can be readily available and work offline, and there is no per inference cost to you or your users.”
Databricks unveiled an agent-context platform stack at Data + AI Summit 2026 Day 1, anchored by Genie ontology and agents.
“We have our agents but they're lacking context. We want to give them all that context on the platform.”
Databricks announces Genie Ontology, which auto-builds a knowledge graph from org data to give agents secure context.
“this is one of the most important things I'm going to announce, Genie Ontology.”
POSETTE 2026 Livestream 4 previews talks on Postgres partitioning, SQL/PGQ graph queries, logical replication, and PG DuckDB analytics.
“SQL PGQ is the latest addition to SQL standard.”
SageMaker AI Async Inference now accepts inline request payloads up to 128,000 bytes, removing the mandatory S3 upload step.
“For payloads up to 128,000 bytes, this removes an entire network round-trip, simplifies client-side code, and reduces the operational surface area of asynchronous inference workloads.”
Postgres' write-ahead log underpins crash recovery, point-in-time recovery, and physical and logical replication.
“Write-ahead log is basically the change-log for the database.”
Postgres CDC via the WAL enables event-driven architectures without dual-write risk, demonstrated across wal2json, Drasi, and Debezium/Kafka.
“In Postgres, every INSERT, UPDATE, and DELETE is already being recorded in the WAL.”
pgcov is a new open-source CLI tool that brings test coverage analysis to PostgreSQL SQL and PL/pgSQL code.
“PostgreSQL code has no coverage visibility yet.”
CloudNativePG's new failoverQuorum feature brings quorum-based consistency to synchronous Postgres replication on Kubernetes.
“if you're running Postgres on Kubernetes, then I do think everybody should use synchronous replication”
AI agents move data retrieval from single-turn RAG to multi-step tool use, but agent-generated SQL often fails retrieval.
“why 90% accuracy in text to SQL is 100% useless”
PostgreSQL 18's planner optimizations (OR-to-ANY rewrite, B-tree skip scan) deliver large real-world query speedups over 17.
“it's like, 1,741 times faster, like 99.9% reduction in execution time”
Serenity Star runs production RAG for AI agents at scale on Azure Database for PostgreSQL.
“It is a live product with thousands of agents and millions of executions, and we use Azure Database for PostgreSQL for our RAG implementation.”
PostgreSQL is becoming the unified 'everything database', replacing fragmented specialized data stacks.
“50% of the developer time, you know, I repeat, 50%, which is pretty much half of the developer time, is wasted in maintaining the pipelines and not really building the new features.”
Local NVMe storage runs Postgres ~100x faster than EBS by eliminating the disk I/O bottleneck behind most performance problems.
“Postgres isn't slow. Your storage is.”
Microsoft built Azure HorizonDB, a 99.9% Postgres database with separated compute/storage and Rust-built storage, for tier-1 scalability.
“HorizonDB is built on the concept of database as logs and separation of compute and storage, which allows compute and storage to scale independently.”
PostgreSQL performance on large tables degrades silently through bloat, vacuum storms, and WAL explosion without proactive maintenance.
“performance doesn't break suddenly, it degrades silently over time”
Amazon Quick adds no-code autonomous agents that work continuously on users' behalf plus a prioritizing activity feed.
“You decide how much autonomy to give each agent, from precise step-by-step instructions to broad goals where agents figure out the path on their own.”
Databricks aims to make enterprise AI work by feeding organizational data, processes, and tacit knowledge as context to frontier and open-source models.
“If we just give the context to these very smart models that can solve those super hard problems, they'll be able to do amazing things inside of our companies.”
Databricks and NVIDIA are partnering to build a full, accelerated AI stack for the agentic era.
Databricks introduces LTAP, a storage technology unifying OLTP and OLAP on one copy of data.
“We think Altab is H-Tab done right and accomplished the goals of H-Tab without actually having a single query engine for it, but we able to actually unify the storage which is by far the most important part.”
Databricks launches Lakehouse RT, a real-time warehouse powered by the Raiden engine running directly on the lake.
“If you have a separate serving stack, test it out and you might be able to collapse the two platforms into one.”
A Microsoft 'Budget Bytes' series teaches building AI apps on Azure SQL for under $25.
“you're going to learn how to build applications with AI and Azure SQL database, all within a budget of $25”