Best of daily.dev — May 2026

1. 1
   Article

   Hacker News·17w

   Im going back to writing code by hand

   A developer spent 7 months vibe-coding a GPU-aware Kubernetes TUI (k10s) with Claude, then archived it after the codebase collapsed under its own weight. The post dissects five concrete failure patterns that emerge from AI-assisted coding without architectural guardrails: AI builds features not architecture (leading to god objects), the god object as default AI artifact, velocity illusion causing scope creep, positional data as a time bomb, and AI mishandling state transitions causing data races. Each tenet includes real code examples from the failed codebase and specific CLAUDE.md/AGENTS.md directives to prevent the same mistakes. The author is rewriting from scratch in Rust, doing architecture design by hand before any AI-generated code.

   414

   43
2. 2
   Article

   Joud Awad·17w

   9/30 Days System Design Question

   A system design challenge presents a Postgres monolith handling 8K writes/min and 40K reads/min where read and write models require fundamentally different data shapes. Four architectural options are evaluated: full CQRS with separate read/write stores, read replicas, denormalizing the write schema, or adding GraphQL with DataLoader. Readers are asked to pick the best approach and justify their reasoning, with the post hinting that one option is a senior engineer trap that works short-term but fails later.

   400

   68
3. 3
   Article

   Joud Awad·17w

   5/30 Days System Design Questions!

   A system design challenge presenting four database sharding strategies for a Postgres orders table at 500M rows with a specific e-commerce workload: 80% customer-centric reads, 15% analytics, 5% writes at 400 RPS. The four options are hash sharding on order_id, range sharding on created_at, directory-based sharding, and consistent hashing with virtual nodes. The post teases that one option is the correct fit and one is a common senior engineer trap, with the full answer promised in comments.

   396

   22
4. 4
   Article

   Dixit Ram·15w

   Mouse Unlock!—no password, just a secret click pattern

   A developer built a mouse-click pattern unlock system for a Linux laptop without biometric sensors. Using Python and the evdev library, a ~150-line daemon reads raw mouse events from /dev/input/, checks lock state via loginctl, and unlocks the screen when a secret click sequence is entered. It runs as a systemd service and works on Wayland. The project was built with Claude Code on Fedora + KDE Plasma.

   463

   99
5. 5
   Article

   Marcos Lobo·18w

   322

   21
6. 6
   Article

   Hackaday·17w

   Running Your Own 3G Network

   1xBTS is a new open-source Rust implementation of the lower three layers of a CDMA2000 (3G) cellular network stack. It supports SDR hardware including USRP B200/B210, LimeSDR Mini 2, and BladeRF Micro 2.0 via SoapySDR. The stack covers BTS, BSC, and MSC layers, supports voice calls via SIP, SMS, and data transfers, and includes a web dashboard with packet tracing. Authentication requires registering handsets in a home location register. Legal caution is advised as cellular band transmissions are heavily regulated.

   311

   6
7. 7
   Article

   Ongoro Blog·19w

   299

   43
8. 8
   Article

   daily.dev World·18w

   daily.dev Hackathon

   daily.dev is hosting a 72-hour async hackathon powered by CodeRabbit, open to all developers. Participants get full access to the daily.dev Public API (feeds, posts, search, bookmarks, profiles, recommendations) and can build across three tracks: Developer Identity, Content Intelligence, and Content → Action. Any tech stack is allowed. Winners receive 1 year of daily.dev Plus and 3 months of CodeRabbit Pro Plus, and get featured on daily.dev. Submissions require a public URL and shareable output shared via social media tagging.

   285

   51
9. 9
   Article

   Joud Awad·17w

   8/30 Days System Design Questions!

   A practical system design scenario involving an e-commerce product catalog running Redis in front of Postgres at 40K RPS. After shipping to production, stale cache data causes customers to see wrong prices and incorrect stock counts. The post presents four cache invalidation strategies — write-through, write-behind, cache-aside, and read-through — and asks readers to identify which pattern best solves the stale data problem given the specific read/write topology described.

   296

   36
10. 10
    Article

    Amit Gajare·18w

    276

    26
11. 11
    Article

    Addy Osmani·16w

    Don't Outsource the Learning

    Addy Osmani argues that the default way developers use AI coding tools optimizes for closing tasks rather than building understanding. Drawing on three recent studies — an Anthropic randomized trial, MIT's 'Your Brain on ChatGPT' EEG research, and a CHI 2026 paper — he shows that passive AI use measurably degrades comprehension and critical thinking. The key finding: the tool doesn't determine the outcome, the posture does. Engineers who used AI to ask conceptual questions outperformed those who copy-pasted generated code. Osmani offers concrete workflow shifts: form a hypothesis before prompting, ask for explanations before code, use learning modes (Claude's Learning Mode, ChatGPT's Study Mode), treat AI output like a junior engineer's PR, and periodically re-derive code by hand. He frames ship velocity and skill growth as two separate metrics — only one of which your employer tracks — and warns that engineers who can only ship with AI are entering a labor market already repricing what expertise is worth.

    265

    19
12. 12
    Article

    Open Source·18w

    Animated icons for React, without the setup pain

    @animateicons/react is a new npm package providing animated Lucide-style icons for React applications. It aims to simplify adding motion details to buttons, menus, dashboards, and landing pages without building custom SVG animations from scratch. Installation is via pnpm/npm and usage follows a straightforward named import pattern from the package.

    261

    26
13. 13
    Article

    The Githubers·15w

    GitHub Copilot is officially DEAD (and my wallet almost died with it)!

    GitHub is transitioning from a flat monthly fee to a usage-based 'AI Credits' billing model starting June 2026, ending the $10/month unlimited plan. A simulation using GitHub's official billing preview tool shows one developer's costs jumping from ~$105/month to ~$2,757/month under the new model — a $2,650+ increase for identical usage. The shift is attributed to the financial unsustainability of flat-rate pricing as AI agents that analyze entire codebases become more common. Heavy users, indie developers, and small agencies are expected to be hit hardest.

    289

    124
14. 14
    Article

    ITNEXT·18w

    The Map of System Topologies

    A comprehensive taxonomy of software system topologies organized by their degree of partitioning into layers (technical) and subdomains (domain). The map covers five major regions: Monolithic systems (true monoliths, monoliths with auxiliary layers, plugins), Layered architectures (ordinary, scaled, and special variants), Plugin family (hexagonal, microkernel, separated presentation), Services area (barebone services, services with extensions, hierarchies), and Fragmented patterns (layers of services, layered services, hierarchies). Each topology is analyzed for its appropriate use case, from small cohesive projects to huge recursive decompositions. Key motifs include managing layers, platform layers, and the principle that system size and complexity drive the choice of partitioning strategy. This is an excerpt from the book 'Architectural Metapatterns: the Pattern Language of Software Architecture'.

    253

    9
15. 15
    Article

    The Githubers·18w

    247

    27
16. 16
    Video

    JavaScript Mastery·19w

    How Senior Engineers Actually Build With AI in 2026 | Build a Full Stack Systems Architecture App

    244

    25
17. 17
    Article

    scopespace·18w

    235

    12
18. 18
    Article

    Joud Awad·16w

    12/30 Days System Design Questions!

    A PostgreSQL table with 200M rows is experiencing write latency spikes from 12ms to 140ms at 8K writes/sec, while dashboard queries scanning 40M rows demand faster reads. The post presents four indexing strategies as a system design challenge: a composite B-tree index, a covering index with INCLUDE columns, routing dashboard queries to a read replica, or a partial index targeting only the hot data slice (signup events from the last 7 days). The key tradeoff is balancing read performance against write throughput when the table already has 4 indexes.

    237

    18
19. 19
    Article

    Bobby Iliev·17w

    221

    3
20. 20
    Article

    Joud Awad·18w

    2/30 Days System Design Questions!

    A classic N+1 query problem is presented: a /orders endpoint fires 51 DB queries per request due to ORM lazy-loading customer data. Four solutions are compared — eager loading with a JOIN (Prisma include), DataLoader batching, Redis caching, and denormalization — each with different trade-offs for future scalability. The post frames it as a team debate exercise rather than providing a definitive answer.

    275

    24
21. 21
    Article

    LeadDev·18w

    What’s gone wrong at GitHub?

    GitHub's reliability has deteriorated sharply, with 257 incidents tracked between May 2025 and April 2026, including 48 major outages. GitHub Actions alone suffered 57 outages in that period. The root cause, per GitHub's CTO, is the explosive growth of agentic AI workflows demanding 30x the platform's designed capacity. High-profile users like Mitchell Hashimoto (Ghostty) and the Zig project have migrated away. Compounding the scaling crisis are engineering failures like an incomplete feature flag that silently reverted thousands of merged pull requests. Microsoft's absorption of GitHub into its CoreAI org and commercial pressure around Copilot are seen as contributing factors. GitHub has declared an 'availability first' mandate, but community patience is running thin as comparable platforms like GitLab and npm handle the same AI-driven growth without comparable disruptions.

    211

    19
22. 22
    Article

    daily.dev World·18w

    213

    50
23. 23
    Article

    daniel.haxx.se·17w

    Mythos finds a curl vulnerability

    Daniel Stenberg, curl's lead developer, shares his experience with Anthropic's Mythos AI model scanning curl's 178K-line C codebase for security vulnerabilities. Despite massive media hype around Mythos being 'dangerously good' at finding security flaws, the scan of curl yielded only one confirmed low-severity vulnerability out of five initially flagged findings — the rest were false positives or non-security bugs. Stenberg notes curl had already been scanned by AISLE, Zeropath, and OpenAI Codex Security, which collectively triggered 200-300 bugfixes. He concludes Mythos is not significantly better than existing AI code analyzers, calling the hype primarily marketing. However, he strongly endorses AI-powered code analysis in general as far superior to traditional static analyzers, urging all projects to adopt such tooling before adversaries exploit unfound flaws.

    208

    8
24. 24
    Article

    Obsidian Blog·17w

    The future of Obsidian plugins

    Obsidian has launched a new Community site and developer dashboard for plugins and themes, replacing the previous GitHub-based submission process. Key highlights include automated security and code quality reviews for every plugin version (not just initial submissions), safety scorecards visible to users, and processing of over 2,300 backlogged submissions. Upcoming features include capability disclosures, verified author badges, private plugins for teams, and multi-collaborator support. Closed-source plugins are no longer accepted for new submissions. The automated review system uses an ESLint plugin and a dashboard preview scan tool, with manual reviews continuing for high-priority cases.

    237

    21
25. 25
    Article

    Better Stack·18w

    206

    17