HiddenMerit Daily · Issue 26

📊 HiddenMerit Daily · Issue 26

Focus on Database Frontiers, Practical Insights for DBAs

May 21, 2026 ｜ 5 Selected Global Breaking News

01｜Google Urges PostgreSQL Developers to Leverage AI Coding at Scale, Alibaba Cloud Leads Domestic Open‑Source Contributions

On May 20, Google published a strong call for PostgreSQL developers to “massively” adopt AI coding tools for code generation, testing, and optimisation during development. This statement comes as Google increases its contributions to open‑source projects like PostgreSQL – earlier this year, Google contributed a series of new PostgreSQL code, including parallel hash join optimisations and new performance monitoring infrastructure. Google’s core argument is that human‑AI collaboration can dramatically reduce monotonous, repetitive tasks in database kernel development, allowing developers to focus on more complex system logic innovation. Since the beginning of this year, the penetration of AI‑assisted coding tools in database kernel development has significantly increased. At the same time, Alibaba Cloud has recently completed large‑scale kernel‑level contributions from PolarDB to the PostgreSQL community, covering key modules such as executor optimisation and lock manager enhancements, demonstrating that the technology backflow path of domestic databases into the open‑source ecosystem is becoming increasingly mature and systematic.

· DBA Perspective: The examples from Google and Alibaba Cloud send a clear signal to DBAs: if you are still only using AI to “write scripts,” your peers may already be using AI to self‑check and generate core operational code. Observing trends, a large portion of PostgreSQL 19 development has already adopted AI‑assisted tools. In the future, a DBA’s core competitiveness will no longer be “memorising syntax and commands,” but the ability to “ask precise questions to AI” and “effectively validate AI output.”

· CTO Perspective: Google integrating AI coding tools into the mainstream development workflow of the database community, along with Alibaba Cloud’s kernel‑level contributions to PostgreSQL, points to the same trend – AI‑driven development is becoming the “new normal” in open‑source foundational software. This is highly significant for enterprises that develop their own data products or engage in secondary development. They should begin exploring how to introduce AI‑assisted coding tools internally to improve iteration efficiency for core software. At the same time, they must establish review and validation mechanisms for AI‑generated code.

· Investor Perspective: AWS, Microsoft, and Google are all heavily investing in AI‑assisted software development capabilities, integrating them into their respective development toolchains. The widespread adoption within the open‑source community confirms this trend. The value growth of technology companies focused on AI coding assistance, automated code review, and test generation is worth continuous attention.

Source: Tech IT168 & Open‑source community disclosures

02｜AWS Redshift Launches Graviton‑Powered RG Instances with Integrated Data Lake Query Engine

In mid‑May, AWS released new Graviton processor‑powered RG instances for its Amazon Redshift data warehouse service, designed to help enterprises manage rising analytics costs and the operational complexity of modern data lake architectures. The core of the new instance is an integrated data lake query engine. AWS stated that this engine can run SQL analytics across both Redshift warehouse data and Amazon S3 data lakes, delivering faster query performance and lower analytics costs – users no longer need to manually move data between multiple systems. At the same time, AWS has recently added several AI‑driven automated management and elastic scaling features to Redshift, further lowering the operational barrier for data warehouses.

Previously, Amazon Web Services also released M7g general‑purpose instances powered by Graviton3, offering 40% better price‑performance than the previous generation, and these have been fully integrated into the RDS managed database service.

· DBA Perspective: Redshift’s move toward the data lake ecosystem means the boundaries of data analytics for DBAs are significantly expanding. In the past, data warehouse DBAs only needed to master Redshift query optimisation and distribution key design. Now, facing a lakehouse‑integrated architecture, they must learn S3 data lake storage formats (e.g., Parquet, ORC) and cross‑engine federated query tuning strategies. When selecting AWS data solutions, the integration of Redshift+EMR+Glue is increasing sharply, and demand for DBAs proficient in cross‑service optimisation will rise notably.

· CTO Perspective: The price‑performance improvement brought by Graviton, combined with the direction of “lakehouse query integration,” directly addresses CTOs’ two main constraints on data infrastructure: cost control and technical complexity. The new RG instance covers both data warehouse and data lake with a single SQL interface, reducing the technical burden of building and maintaining two separate engines within an organisation. For enterprises already deeply using AWS, it is recommended to pilot some analytical workloads on the new instances and quantify the cost‑benefit ratio.

· Investor Perspective: AWS embedding data lake query capability into Redshift signals that leading cloud data warehouse vendors are further encroaching on the feature territory of traditional data lake solutions. Investors should be cautious of data lake startups that lack independent technical moats and overly rely on a “connector model.” At the same time, pay close attention to whether vendors marketing “Lakehouse” can find differentiation under the pressure of AWS and Microsoft.

Source: IT168 & AWS official announcements

03｜Oracle APEX 26.1 AI Agent Security Tested Within Days, Claude Successfully Defends 70% of Attacks

On May 14, Oracle APEX 26.1 was GA. Three days later, a security engineer successfully deployed a full containerised environment of Oracle AI Database Free 23.26.1.0.0, APEX 26.1.0, and ORDS 26.1.1, and integrated the new AI Agent + Tools functionality with the Anthropic Claude Sonnet 4.6 model. In the subsequent 72‑hour deep adversarial red‑team test, the tester launched 10 jailbreak injection attacks against the AI Agent. The results showed that Claude autonomously defended against 7 of those attacks – the AI engine identified and blocked malicious intent in SQL generation or instruction execution without any additional pre‑set protection rules, with only a few attacks slipping through. This outcome indicates that, in low‑code AI applications, the security alignment capability of large models is beginning to serve as an effective “dynamic defence line.” In the future, the security foundation of AI agents will rely on a dual‑track mechanism of “human pre‑set rules” and “model‑intrinsic defence.”

· DBA Perspective: The adversarial test results of the APEX AI Agent offer an important lesson for DBAs: database access security is moving from “static privilege management” to “dynamic behaviour monitoring.” As AI agents take on more SQL generation and execution tasks, DBAs will face new dimensions such as “AI output auditing” and “establishing model behaviour baselines.” It is advisable to learn in advance about RAG jailbreak prevention techniques and LLM output security validation methods to prepare for the routine operations scenarios where AI agents “reside” in databases at scale.

· CTO Perspective: The security validation of Oracle’s low‑code platform integrated AI Agent shortly after its release reflects enterprises’ high alert regarding low‑code AI safety. Claude’s ability to autonomously defend against 70% of attacks without additional protection suggests that model supply chain security can serve as a defence layer, but this does not mean enterprises can rely solely on AI. A multi‑layer protection mechanism of “model‑intrinsic security + human rules + runtime monitoring” should be established.

· Investor Perspective: Oracle’s APEX AI Agent adversarial test provides the market with an important sample: the security capability of AI agents is becoming a key threshold for enterprise‑grade AI application deployment. Security companies that provide runtime security monitoring for AI agents, adversarial testing services, and attack simulation tools will face new demand growth.

Source: Dev.to & Oracle tech community

04｜Domestic Databases Accelerate Deep AI Integration and Ecosystem Expansion: TencentDB Upgrade, ESG Multi‑Dimensional Database Released

On May 20, the 2026 Tencent Cloud Convergence Innovation Summit was held in Beijing. The new version of the TDSQL database was officially released, featuring a deeply refactored enterprise‑edition compute engine with OLTP performance improved by 50%, OLAP by 20 times, 99% centralised syntax compatibility, and a basic edition 100% compatible with MySQL. On May 14, Tencent Cloud had already announced the open‑sourcing of TencentDB Agent Memory, providing short‑term memory compression and long‑term personalised memory capabilities for agent long‑task scenarios. The summit further strengthened a “one cloud, multiple models” full‑process AI service system. On the ecosystem cooperation front, Desheng Technology and CETC Kingware signed a strategic cooperation agreement in Guangzhou on May 19, focusing on the human resources and healthcare sectors, as well as data element application scenarios. They reached agreements on AI application implementation and industry solution creation, driving digital transformation of livelihood services. Kingware KES V9 2025 Fusion Edition has demonstrated significant performance gains in core system replacement in the energy and telecommunications sectors – under high‑concurrency billing and energy dispatch data loads, core transaction response time was reduced by an average of 40%, and system throughput increased by 30% to 40% on the same hardware configuration. Kingware’s native multi‑modal convergence architecture supports unified storage and querying of eight data models including relational, vector, graph, document, time‑series, and spatial.

Following the strategic cooperation signing between Desheng Technology and CETC Kingware, the two parties will jointly promote large‑scale implementation of domestic databases in human resources, healthcare, and other scenarios. Additionally, database independent software vendor ESG Data Systems recently released its new multi‑dimensional spatio‑temporal database V1.0, targeting emerging AI scenarios such as connected vehicles, smart cities, and financial risk control, providing fused analysis capabilities for time‑series and spatial data, further complementing the product matrix of domestic databases in the multi‑modal convergence direction.

· DBA Perspective: The 20‑fold OLAP performance improvement in the upgraded TDSQL version is a major leap toward accelerating lakehouse and vector workloads. DBAs should reassess the cost‑effectiveness of migrating existing analytical workloads to the new version, while also paying attention to the actual performance of transaction and analytical workloads under complex mixed loads. The real‑world data of Kingware KES V9 achieving a 40% response time reduction and 30%‑40% throughput increase in energy and telecom sectors provides a high‑value practical reference benchmark for DBAs making financial Xinchuang technology selections.

· CTO Perspective: Tencent Cloud’s emphasis on convergence innovation capabilities in the agent era, combined with Agent Memory open‑sourcing and TDSQL engine upgrades, builds a closed loop for domestic data intelligence covering development, operation, and governance. The cooperation between Kingware and Desheng Technology targets critical livelihood scenarios such as human resources and healthcare, indicating that domestic databases are advancing from “financial replacement” to “core social infrastructure.” It is recommended to pay attention to the performance of the new TDSQL version under complex mixed loads and use the open‑sourced Agent Memory tools to build intelligent data application pilots, gradually verifying the technical maturity of full‑stack domestic database replacement.

· Investor Perspective: The recent intensive actions by leading domestic database vendors – Tencent Cloud, Kingware, Dameng – in AI integration and industry ecosystems mark the industry’s entry into a stage of comprehensive intelligence and scenario‑focused deep cultivation. At the capital level, enterprises with “in‑house kernel development + deep AI integration + industry solution delivery” capabilities are more likely to receive long‑term increased holdings from institutional investors. It is recommended to track the commercial deployment progress of the new TDSQL version in the financial industry, as well as the pace of Kingware’s large‑scale penetration in key areas such as energy and human resources.

Source: Tencent Cloud Summit, Taiji Computer announcement, Desheng Technology official WeChat, Kingware tech blog & ESG release

05｜Russian Telecom Giant Partners with SberTech to Launch Domestic DBMS, Replacing Oracle and SQL Server with Pangolin DB

On May 20, 2026, Russian telecom giant Rostelecom and SberTech announced a partnership to adopt the Russian self‑developed Pangolin DB database management system to replace foreign database solutions such as Oracle Database and Microsoft SQL Server in critical information systems. This move is a significant implementation of Russia’s national import substitution strategy in infrastructure software, aimed at enhancing national data sovereignty and the security and controllability of critical information infrastructure. Pangolin DB will be gradually deployed into Rostelecom’s core telecom business systems and SberTech’s financial data processing platforms, with a phased migration plan toward full domestic replacement. The two parties also confirmed they will jointly develop industry‑specific data platforms based on Pangolin DB and promote them to other Russian government and commercial institutions.

· DBA Perspective: The global “de‑Oracleisation” wave has spread from Europe and America to Russia, reinforcing the trend that the traditional skill moat of Oracle DBAs in non‑Chinese regional markets is rapidly narrowing. However, the domestic DBMS ecosystem in countries like Russia is still under construction, offering a new technology export window for DBAs skilled in distributed architecture tuning and multi‑modal data management.

· CTO Perspective: The joint action by Rostelecom and SberTech confirms that requirements for “data sovereignty” and “autonomous control of critical systems” are rising to the level of national strategy. When planning their core technology stacks, enterprises are advised to incorporate supply chain security and vendor diversification into long‑term considerations, and appropriately increase the evaluation and reserve of open‑source and geopolitical backup solutions.

· Investor Perspective: Russia’s domestic replacement solution through Pangolin DB reflects that “data sovereignty” is becoming an important arena in global technology competition. This signals that more regions will increase their development and procurement of domestic databases, opening up new geographic incremental markets for database enterprises with independent core technology. At the same time, the market share of traditional giants such as Oracle will face further pressure globally.

Source: AKM.RU & Russian domestic software replacement plan disclosure

📅 Recent Database Hot Topics Recap

Date Event Core Highlights

May 19 Desheng Technology and CETC Kingware sign strategic cooperation agreement Focusing on human resources and healthcare, driving digital transformation of livelihood services and large‑scale domestic database implementation

May 19 Russian Rostelecom and SberTech adopt Pangolin DB Domestic DBMS replaces Oracle and SQL Server; significant implementation of national import substitution strategy

May 20 Google urges PostgreSQL developers to adopt AI coding at scale AI‑assisted development penetrates database kernel; Alibaba Cloud completes large‑scale PolarDB contributions to PG community

May 20 2026 Tencent Cloud Convergence Innovation Summit & TDSQL new version release TDSQL new version: OLTP +50%, OLAP +20x; Agent Memory open‑sourced, further improving agent memory foundation

May 20 ESG releases multi‑dimensional spatio‑temporal database V1.0 Targeting connected vehicles, smart cities, financial risk control; strengthening domestic DB multi‑modal convergence matrix

May 20 Kingware core system replacement in energy & telecom sectors Core transaction response time reduced by 40% on average; throughput increased 30%‑40%; validating domestic replacement capability in critical scenarios

Mid‑May AWS Redshift launches Graviton‑powered RG instances Integrated data lake query engine unifies S3+Redshift analytics entry; improved price‑performance

Mid‑May Oracle APEX 26.1 AI Agent security test Claude autonomously defends 7 of 10 attacks; AI agent “intrinsic defence” effectiveness begins to show

May 22 XCOPS Intelligent O&M Managers Annual Conference (Guangzhou) Focusing on large‑model application practices, intelligent O&M, and database technology frontiers

May 29 Tencent Cloud “Database + AI” product launch Debut of six core AI‑In‑Database engines; data foundation for the agent era officially unveiled

📌 Issue Summary

News Core Keywords DBA Actions CTO/Decision‑Maker Focus Investor Perspective

Google urges PostgreSQL AI coding AI‑assisted development, PostgreSQL kernel contributions, PolarDB code backflow Shift from “memorising syntax” to “asking AI questions + validating output” Introduce AI‑assisted coding tools internally; improve core software iteration efficiency Value of AI‑assisted software development toolchain companies continues to grow

AWS Redshift Graviton instances Data lake query engine, Graviton price‑performance, lakehouse integration Learn S3 data lake storage formats and cross‑engine federated query tuning Use new instances to reduce costs and improve efficiency; pilot analytical workload migration Cloud data warehouse leaders encroach on data lake feature territory; caution on “connector‑mode” startups

Oracle APEX AI Agent security test AI Agent jailbreak injection, Claude intrinsic defence, dynamic security line Learn RAG jailbreak prevention and LLM output security validation; prepare for AI Agent operations Build multi‑layer protection: model intrinsic security + human rules + runtime monitoring AI Agent security capability becomes threshold for enterprise‑grade application deployment

Domestic DB AI integration & ecosystem expansion TDSQL performance boost, Agent Memory open‑sourced, Kingware energy/telecom results Evaluate new version mixed‑load price‑performance; use Kingware real‑world data for selection Build intelligent data application pilots with TDSQL new version + Agent Memory Enterprises with “in‑house kernel + deep AI + industry solutions” favoured by capital

Russian Pangolin DB replaces Oracle Data sovereignty, domestic DBMS, import substitution Focus on distributed architecture tuning opportunities for export to multiple markets Incorporate supply chain security and vendor diversification into long‑term planning Traditional giants’ global market share faces further pressure

HiddenMerit Team Production

Slogan: 绩优隐于内，金石启新程 | Hidden deep. Merit bold. Forge ahead.