sonichigo

When we think about CI/CD, the database is often a blind spot. I wrote this post to explore why this is so risky and what makes database changes so different from application code. The article dives into the unique challenges, from black-box migrations to performance regressions—and how we can begin to address them by shifting observability left!

We're validating our approach for database GitOps and noticing that traditional branching-per-environment models often introduce significant friction:

1. Merge conflicts across dev, qa, prod branches 2. Inconsistent baselines and forgotten hotfixes 3. Broken audit trails and configuration drift

We shifted instead to a trunk-based approach with Liquibase contexts:

1. All changes stored in one main branch 2. Environment-specific behavior driven by context metadata 3. Promotions handled via pipelines, not merges 4. Full GitOps traceability and one-click rollback support

We’ve been reevaluating our Git strategy for database changes. Historically, our team maintained one branch per environment—dev, qa, and prod. While it gave us some flexibility, it created a lot of overhead: merge conflicts, forgotten hotfixes, and inconsistent deployment behavior.

Recently, we adopted a trunk-based model paired with Liquibase for change management. Instead of branching per env, we:

Keep all changelogs in a single main branch Use Liquibase contexts to control env targeting Promote via automated pipelines Apply GitOps principles for traceability and rollback

Curious—how are others approaching Git branching for DB changes? Is anyone still using GitFlow or feature/env branches? Anyone managing this at scale with other tools like Flyway or Atlas?

Would love to hear real-world stories.

Like many teams, we treated application CI/CD as a first-class citizen but left PostgreSQL schema changes as a manual afterthought. That led to broken deployments, inconsistent environments, and rollback nightmares.

This article shares our transition to a fully automated, GitOps-style workflow using Harness Database DevOps and Liquibase. It dives into technical challenges, architecture changes, and lessons learned while scaling PostgreSQL migrations across multiple environments.

Read the full blog post - https://dub.sh/JMUgbep

Would love to hear how others have tackled this problem at scale.

Thanks for the feedback, I'll work on to fix this and update it

DeepCoder-14B is quietly outperforming some of the biggest coding LLMs out there—including o3-mini—and it’s open source from top to bottom.

Find out how a lean 14B model, trained on real-world problems and refined with smart RL techniques, is changing the game.

the tool Looks interesting. What all metrics are been tracked to gauge the intent of the visitor?

I recently wrote a fascinating blog post, "The Marvelous Rise of AI Test Case Generation", which dives into how AI is revolutionizing software testing by automating the creation of test cases.

The article touches on:

- How AI tools are learning from production data to generate edge-case scenarios. - The impact of AI on traditional testing practices and its potential to enhance developer productivity. - Challenges and limitations of adopting AI for this purpose.

I’m curious to hear from the HN community:

1. Have you tried any AI-driven test case generation tools? If so, which ones, and how effective were they? 2. What’s your take on the future of AI in the testing space? 3. Are there specific challenges or gaps in AI testing solutions that still need to be addressed?

I mean Apple's support period is actually a minimum of five years of software support, which gives the company some leeway when it comes to actually shipping updates. But still Apple's behindWhile that could change with the iPhone 16 series this fall, let's see how it goes.

I'll update it thanks for pointing

In COMPUTEX Nvidia's CEO also mentioned that, companies and countries are partnering with NVIDIA to shift the trillion-dollar traditional data centers to accelerated computing and build a new type of data center — AI factories — to produce a new commodity: artificial intelligence

> Refering - https://news.ycombinator.com/item?id=40593858

The new Copilot Extensions feature in GitHub Copilot, according to GitHub senior vice president Mario Rodriguez, has been designed to accelerate developer workflows without users having to exit the GitHub window or their IDEs, similar to VSCode extensions.

During COMPUTEX, CEO Jensen Huang, announced that big shot companies like ASRock Rack, ASUS, GIGABYTE, Ingrasys, Inventec, Pegatron, QCT, Supermicro, Wistron and Wiwynn will deliver cloud, on-premises, embedded and edge AI systems using NVIDIA GPUs and networking.

At Computex 2024, March, AMD introduced its latest generation of Ryzen laptop processors designed for generative AI workloads called - the Ryzen AI 300 series. Seems like more and more companies are shifting from NVIDIA for AI model computation.

that's the thing tho, today grades aren't much of relevance since skills are more favoured but there's a lot of competetion as well

Since these are artifical neurons, i would say there's a 70-30 chance of above happening.

one of the best i have ever seen. Sadly it didn't recieved attention it deserved earlier but is getting popular over socials now a days a lot