A Better Workflow for AI-Assisted Coding: Lessons from Matt Pocock
I recently watched Matt Pocock's workshop, Full Walkthrough: Workflow for AI Coding . The main takeaway: Good AI-assisted coding is not about writin…
Architecture
⚑ Report a ProblemLatest Architecture news from Tech News
All topics
agents
ai
api
architecture
automation
aws
backend
beginners
career
database
devchallenge
devops
javascript
llm
machinelearning
mcp
news
opensource
performance
productivity
programming
python
react
security
showdev
softwareengineering
systemdesign
tutorial
typescript
webdev
The Overlooked Reason Your RAG Pipeline Keeps Returning Garbage
I spend a lot of time in the AI space -- reading papers, building things, talking to engineers who are actually shipping. And there is a gap between…
The Exact Stack I Use to Build Production AI Agents (No Fluff)
I spend a lot of time in the AI space -- reading papers, building things, talking to engineers who are actually shipping. And there is a gap between…
I built the world's fastest AI gateway in Rust
TL;DR I built Antirion , an open-source (Apache-2.0) AI gateway written in Rust: one binary between your apps and LLM providers, with routing, fallb…
Синтетическая эмпатия: как RAG, системные промпты и TTS инженерят привязанность к LLM-агентам
ELIZA в 1966-м заставляла секретарш просить не заглядывать в распечатки диалогов. Шестьдесят лет спустя компаньон-агенты на LLM делают то же самое,…
Version the wire schema, normalize at the boundary, and keep failures visible
Problem statement When you change a function-calling or Structured Outputs schema after users have active conversations, older turns and replayed to…
Write your agent as functions first, add a framework later
Plenty of agent tutorials open with a framework import. I'd start with ordinary functions that call a model, return structured data, and invoke tool…
Pick the LLM eval metric before you build the dashboard
You can ship a polished eval dashboard and still learn nothing about whether the product works. A common way to get there is picking a number becaus…
Claude Sonnet 5.5 and Haiku 5.5 in Python: Two SDKs, Prompt Caching and Haiku's 100K Pricing Cliff
Short answer: call claude-haiku-5-5 for classification, extraction, routing and summaries, and claude-sonnet-5-5 for multi-step coding and long tool…
How LLMs are actually built — and why most "how it works" threads stop too early
Most "how LLMs work" explainers end at transformers and attention. The parts that decide whether a model is actually good — the data pipeline, the p…
How to Install ArcLM, Tokenize Text, and Run a Tiny Language Model
How to Install ArcLM, Tokenize Text, and Run a Tiny Language Model Want to experiment with language models in Python without starting by downloading…
Durable execution keeps the state. It doesn't keep the rules.
Originally published at zarel.ai . The strongest infrastructure story in agents right now is durable execution: an agent can run for hours or days,…
Frontier Pre-Training, AI Hardware Hacking, and Legal Precedents
Today's AI landscape is shifting toward the infrastructure of intelligence. From Microsoft's deep dive into frontier pre-training to new legal rulin…
AtmoRouter: A Cheaper OpenRouter Alternative (OpenAI-Compatible, Pay-Per-Token)
OpenRouter made a great idea mainstream: one API, many models. Subscription coding tools like Command Code bundle model credits into monthly plans.…
Can decision models like Jev and Laya route a coding agent?
We plugged Laya, a 421M-parameter decision model, into our coding agents as a model router. The plumbing worked in about 112 ms. Deciding when a big…
Semantic Caching Explained: When It Saves Money and When It Fails
TL;DR Semantic caching evaluates the mathematical similarity of prompt embeddings against stored responses to bypass redundant upstream model calls.…
Can AI Write Verilog for FPGAs
Can AI Actually Write Verilog for FPGAs? My Real-World Test As someone who does both full-stack web development and FPGA/Verilog programming, I was…
Coinbase's 9B Fraud Model Beat Opus 4.5: Copy the Pattern With a Verifiable Reward, Tier-Only Output, and Block-Only Harness
Attributed compile + one small offline run on synthetic data Primary sources (Coinbase Engineering, "Owning Intelligence" series): Part 1/3 — Build…
Three models vote on what happened
I ran the same job past three different AI providers and had them vote on the answer, and I did it for a hobby project about a Sunday night D&D…
Japan's devs on Claude Code: nerve, not capability
Originally published in TL;JP , a weekly, fact-checked read on what Japanese AI and tech practitioners are saying on X. This is the Oct 6-12, 2026 i…
I Don't Want Another ChatGPT. I Want a Private Chief of Staff: The Architecture of a $1M Personal AI Agent
Most AI products are built for companies. They optimize dashboards, workflows, compliance, and org-wide productivity. But there is another market th…
Зачем нам 200 инструментов и 5 чатов: как мы строим ИИ-агентов в финтехе
Сегодня интеграцией искусственного интеллекта в рабочие процессы никого не удивишь. Однако разница между быстрым прототипом на базе стороннего API и…
SubQ Snap: The Hype, the Benchmarks, and the Reactions
Seven points and one comment. That is what the SubQ Snap launch got on Hacker News, hours in. The comment was a question, and it was not friendly: D…
Claude Account Banned? Practical Alternatives and What to Do Next (2026)
Originally published at 51发现导航 (51 Discovery Nav) . Losing access to Claude in the middle of a project is frustrating, especially if you relied on i…
The Agent That Trapped Itself: How Context Eviction and Deduplication Broke on a Go Project
The Agent That Trapped Itself: How Context Eviction and Deduplication Broke on a Go Project A deep dive into a real failing run in the Legion multi-…
My Triton Kernel Ties bitsandbytes — and Both Lose to Math
A fused W4A16 GEMV for batch-1 decoding, measured against the RTX 3060 roofline. Scope note: one RTX 3060 12GB, batch-1 greedy decoding, Qwen2.5-1.5…
Как устроен агент: архитектура поверх LLM
Про архитектуру трансформеров написано уже немало, поэтому опишу работу LLM агента с прикладной точки зрения. Читать далее
Ragas vs Truvec for RAG Retrieval Evaluation
I have built and debugged many RAG systems, and the most common failure I saw was not because of bad prompts or weak LLM models. Most failures were…
Why I Stopped Predefining Agent Roles — and Built a Runtime That Grows Its Own Team
Most multi-agent frameworks ask you to design the team before you know the task. You define roles (researcher, writer, reviewer), wire up a workflow…
Four numbers Ollama 0.40 gives you now, and which ones to trust
Ollama went from 0.33 to 0.40 in six weeks, and most of what changed is invisible unless you read the responses with a magnifier. I did, because I m…