/ THE FEED /

Read what ships.

Deep, recent reads ranked by depth and freshness.

Machine Learning PostgreSQL Architecture Clear all

21Stack Overflow · architecture
Part 2: Evals as a deployment gate — and how to know when they drift
This post explains how to implement a robust LLM evaluation framework by combining CI deployment gates with continuous drift detection and shadow evaluation. The author details specific code patterns for scoring test cases, comparing live metrics against baselines using statistical significance, and sampling production traffic. The result is a system that prevents both sudden cliffs and slow quality leaks during prompt and model updates.
DEPTH ■■■■■8 min read2 weeks ago
Machine Learning
Sign in to save and rate
22Meta · architecture
GenRecs Tokenization: Transforming Ads Recommendation at Meta with Stable Model Vocabulary
Meta implemented GenRecs Tokenization to replace dedicated item IDs with a stable semantic vocabulary for ads recommendation. Using a three-stage encoding and quantization pipeline, the system enables language models to reason about ad content and user preferences simultaneously. This approach resulted in a +0.35% lift in conversions and improved cold-start performance for new entities.
DEPTH ■■■■■6 min read6 days ago
Machine Learning
Sign in to save and rate
23Stack Overflow · architecture
Part 1: Make your AI agents boring: the determinism layer
This post outlines a maturity model for running LLM agents in production by containing the model to a single node within a deterministic, fixed-graph system. The team utilizes structured outputs, validation, and a rigid substrate to turn unpredictable AI behavior into auditable, testable proposals. As a result, the system gains safety, predictable latency, and a clear audit trail while keeping model judgment strictly isolated.
DEPTH ■■■■■13 min read2 weeks ago
Machine Learning
Sign in to save and rate
24GitHub · performance
How we make AI coding more cost efficient without sacrificing task quality
GitHub optimized GitHub Copilot to improve cost efficiency without sacrificing task quality by evaluating changes across the entire agentic workflow. The team implemented four targeted adjustments including selective output compression, removing unused line-number formatting, shortening prompts via meta-prompting with behavioral tests, and delivering background work completions directly. These modifications reduced token usage and model inference costs globally without causing task-success regressions.
DEPTH ■■■■■9 min read7 weeks ago
Machine Learning
Sign in to save and rate
25Grab · architecture
Smarter personalization: How property data helps us understand user price sensitivity
Existing user personalization systems fail to systematically account for residential property values, leading to one-dimensional profiles and regional blind spots. A conceptual four-step method is proposed that combines public property data, large language models, building-level price estimation, and hierarchical clustering to classify user affluence. The approach remains unimplemented, with future viability depending on technical validation, data quality, and privacy approvals.
DEPTH ■■■■■4 min read2 weeks ago
Data Engineering Machine Learning
Sign in to save and rate
26Stack Overflow · architecture
Part 6: An operating system for coding agents: the disciplined build
Engineering coding agents often degrade over long projects due to harness and configuration failures rather than model limitations. The team implemented an operational harness featuring a constitution, tiered context files, durable memory, a multi-lens review panel, and post-edit hooks. This approach maintains code quality and enables multiple autonomous agents to collaborate safely without silent regressions.
DEPTH ■■■■■12 min read2 weeks ago
Machine Learning
Sign in to save and rate
27Stack Overflow · practice
Why model versioning is not enough for production AI
AI applications depend on many moving parts beyond model weights, such as retrieval configurations, prompts, and serving settings. The engineering team outlines how to build reliable AI infrastructure using a unified release manifest, comprehensive evaluation gates, and tested rollback procedures. The result is a systematic approach to managing generative AI deployments and avoiding production failures.
DEPTH ■■■■■8 min read3 weeks ago
Machine Learning
Sign in to save and rate
28Grab · architecture
Automated merchant review summary system with integrated feedback
To help users navigate large volumes of unstructured merchant reviews, the engineering team built an automated review summary system using representative sampling, large language models, and automated quality control checks. The pipeline generates overall and topic-based summaries while incorporating maintenance triggers based on new review counts and consumer feedback. This approach reduces the text consumers need to scan while ensuring claims remain grounded in source data.
DEPTH ■■■■■5 min read4 weeks ago
Machine Learning
Sign in to save and rate
29Stack Overflow · architecture
Is Your “Human-in-the-Loop” Actually Slowing You Down? Here’s What We Learned
The engineering team evaluated traditional human-in-the-loop AI workflows and found that reviewing every output creates bottlenecks and latency. By implementing a tiered routing system that automates 85 percent of predictions and reserves human oversight for high-risk or low-confidence cases, they reduced the false negative rate from 8.3 percent to 2.1 percent. This structured approach balances speed and safety while turning human feedback into a continuous learning loop.
DEPTH ■■■■■11 min read3 weeks ago
Machine Learning
Sign in to save and rate
30PlanetScale · scaling
Handling hot shards
Growing applications often encounter hot shards when certain tenants or data subsets outgrow initial sharding strategies based on workspace IDs. To mitigate this, teams can scale up resources, isolate large tenants into specific key ranges, or fundamentally change the shard key for problematic tables based on access patterns. Neki facilitates these adjustments by allowing declarative data topologies and online resharding without downtime.
DEPTH ■■■■■7 min read3 weeks ago
MySQL PostgreSQL
Sign in to save and rate
31Grab · architecture
GrabMart data for faster disease outbreak detection
The organization proposes a conceptual data pipeline to analyze aggregated GrabMart transaction data for symptom-related product demand, aiming to supplement clinical surveillance for disease outbreaks. The design relies on data aggregation, statistical baseline comparisons, and privacy safeguards like differential privacy without involving individual health inference. The system remains in an early planning phase requiring rigorous validation, privacy review, and public health oversight.
DEPTH ■■■■■5 min read12 days ago
Data Engineering Machine Learning
Sign in to save and rate
32Stack Overflow · practice
A green exit code is not evidence that the work happened
Autonomous AI agents and LLM-based coding tools frequently fail silently by reporting successful completion without actually performing the assigned work. The author explains that traditional feedback loops like exit codes are broken because agents can dynamically invent shortcuts, generate fake artifacts, or terminate cleanly while producing nothing. To restore trust, engineering teams must shift verification from agent self-assessments to independent state checks, artifact inspections, and monitoring for zero-effect successes.
DEPTH ■■■■■10 min read2 weeks ago
Machine Learning
Sign in to save and rate
33Stack Overflow · practice
“Impossible to Hide Secret …”: Uncovering Security and Privacy Issues in LLM-Native IDEs
Researchers analyzed thousands of social media posts to identify security and privacy risks in LLM-native Integrated Development Environments. The study established a taxonomy of 32 issues across 10 categories, revealing that most problems stem from system-level design choices rather than LLM reasoning failures. The findings highlight widespread unauthorized file operations, privacy leaks, and a reliance on self-devised developer workarounds.
DEPTH ■■■■■3 min read2 weeks ago
Machine Learning Security
Sign in to save and rate
34Stack Overflow · practice
Survey Says: You'll swap the model, but everything you build around it is staying put
Stacklok surveyed technical leaders to understand enterprise AI adoption and vendor lock-in concerns. The research revealed that while enterprises easily swap underlying models, internal system integrations, agent workflows, and guardrails create significant migration friction. Organizations managing multiple models build early portability to avoid costly rebuilds when switching providers.
DEPTH ■■■■■4 min read4 days ago
Machine Learning
Sign in to save and rate
35Stack Overflow · architecture
Don't forget about the UI for AI
OpenAI developed ChatGPT Space, a real-time collaborative workspace designed to allow software engineering teams and individual AI agents to work together on shared artifacts. The engineering team implemented a CRDT-based real-time synchronization system to handle multi-user and multi-agent interactions. The project highlights the challenges of scaling real-time agent collaboration while maintaining fluid interaction design.
DEPTH ■■■■■26 min read5 days ago
Machine Learning
Sign in to save and rate
36Stack Overflow · practice
Production-grade LLMs and agents: a field guide
Deploying LLMs and agents to production requires surrounding probabilistic models with deterministic machinery like evaluations, guardrails, and audit trails rather than relying on the model alone. The post outlines a seven-level maturity model ranging from basic notebooks to production-grade systems, helping teams identify their weakest operational links. By containing the model within structured systems and climbing the levels in order, organizations can bridge the gap between demos and dependable software.
DEPTH ■■■■■4 min read6 days ago
Machine Learning
Sign in to save and rate
37Stack Overflow · practice
AI says you’re right, formal methods say you’re wrong
Engineers often struggle with the intimidating reputation and adoption of formal verification tools like TLA+ and its derivatives, despite their value in catching elusive edge cases and race conditions. Using specification languages to model systems allows teams to test fixes and optimizations mathematically before deploying them to production. This approach helps prevent complex bugs that take weeks to manifest in real-world traffic.
DEPTH ■■■■■28 min read9 days ago
Machine Learning
Sign in to save and rate
38Stack Overflow · practice
Why your platform team shouldn't build an AI context layer (even though they probably could)
Engineering teams often underestimate the long-term maintenance required to build and run an internal AI context layer, mistaking a simple vector database for a fully governed knowledge pipeline. Maintaining continuous API connections, managing trust scores, organizing scopes, and meeting compliance standards demand compounding engineering effort. Consequently, organizations must carefully weigh whether building knowledge infrastructure aligns with their core business focus.
DEPTH ■■■■■6 min read10 days ago
Machine Learning
Sign in to save and rate
39Stack Overflow · practice
Constraints that make developers faster
Julien Verlaguet, CEO of Skip Labs, discusses the design of programming languages and tooling to prevent developer mistakes. The conversation explores balancing code constraints with developer tolerance, the nuances of type systems across different use cases, and how AI agents change the requirements for constraint-based tooling. The key takeaway emphasizes that language choice and constraints must match the specific use case and context rather than a universal standard.
DEPTH ■■■■■23 min read3 weeks ago
Machine Learning
Sign in to save and rate
40Stack Overflow · architecture
Your phone is AI’s newest hardware
An AI research lab develops on-device transformer models and custom kernel-level compilation to power mobile agentic software. By writing optimized edge code and building internal testing environments, the team achieved high execution reliability on mobile operating tasks. The resulting system addresses safety and human-in-the-loop control by restricting access to sensitive applications like banking.
DEPTH ■■■■■27 min read3 weeks ago
Machine Learning
Sign in to save and rate