A minimal, elegant blog exploring modern web development, architecture, and design.https://soumitrasaha.com/en-ushttps://soumitrasaha.com/blogs/how-i-think-about-ai-native-software-development/https://soumitrasaha.com/blogs/how-i-think-about-ai-native-software-development/AI Native is not just adding a chat box. It is designing the product, codebase, and feedback loop around what models can reliably do.Thu, 03 Sep 2026 18:30:00 GMTAI NativeAgentsArchitectureSoftware Engineeringhttps://soumitrasaha.com/my-reads/anthropic-claude-opus-48-agent-reliability/https://soumitrasaha.com/my-reads/anthropic-claude-opus-48-agent-reliability/Anthropic's Opus 4.8 release is worth reading for its focus on judgment, honesty, tool use, and long-running agent work.Thu, 03 Sep 2026 18:30:00 GMTAnthropicClaudeAgentsCodinghttps://soumitrasaha.com/my-reads/arena-text-leaderboard-human-preference-signal/https://soumitrasaha.com/my-reads/arena-text-leaderboard-human-preference-signal/Arena is worth checking because blind human votes catch qualities that static benchmarks often miss, especially open-ended writing and reasoning.Thu, 03 Sep 2026 18:30:00 GMTModel ComparisonsArenaAI ModelsEvaluationhttps://soumitrasaha.com/my-reads/artificial-analysis-model-price-provider-comparison/https://soumitrasaha.com/my-reads/artificial-analysis-model-price-provider-comparison/Artificial Analysis is useful because it compares models across quality, speed, latency, price, context, and provider behavior instead of only one leaderboard score.Thu, 03 Sep 2026 18:30:00 GMTModel ComparisonsArtificial AnalysisAI ModelsBenchmarkshttps://soumitrasaha.com/my-reads/baidu-ernie-51-efficient-multimodal-scaling/https://soumitrasaha.com/my-reads/baidu-ernie-51-efficient-multimodal-scaling/ERNIE 5.1 is worth tracking because Baidu is talking about model compression, elastic pre-training, agentic post-training, and cost-performance.Thu, 03 Sep 2026 18:30:00 GMTBaiduERNIEChinese AI LabsMultimodalhttps://soumitrasaha.com/my-reads/cloudflare-agent-development-lifecycle/https://soumitrasaha.com/my-reads/cloudflare-agent-development-lifecycle/A sharp Cloudflare article about why agent-driven software needs APIs, previews, observability, and lifecycle design beyond the normal SDLC.Thu, 03 Sep 2026 18:30:00 GMTCloudflareAgentsSDLCDevOpshttps://soumitrasaha.com/my-reads/cloudflare-ai-gateway-workers-ai-control-plane/https://soumitrasaha.com/my-reads/cloudflare-ai-gateway-workers-ai-control-plane/A useful Cloudflare read about treating AI calls as traffic that needs routing, logs, billing, retries, and control from the start.Thu, 03 Sep 2026 18:30:00 GMTCloudflareAI GatewayObservabilityModel Routinghttps://soumitrasaha.com/my-reads/cloudflare-browser-run-for-agents/https://soumitrasaha.com/my-reads/cloudflare-browser-run-for-agents/Browser Run is useful because real agents need browser control, live view, session replay, CDP access, and human handoff when the web gets messy.Thu, 03 Sep 2026 18:30:00 GMTCloudflareBrowser RunAgentsAutomationhttps://soumitrasaha.com/my-reads/cloudflare-internal-ai-engineering-stack/https://soumitrasaha.com/my-reads/cloudflare-internal-ai-engineering-stack/This is a strong read on how a large engineering org actually uses AI coding tools, standards, review, and platform primitives together.Thu, 03 Sep 2026 18:30:00 GMTCloudflareAI EngineeringDeveloper ToolsPlatformhttps://soumitrasaha.com/my-reads/cloudflare-kitesurf-agent-browser/https://soumitrasaha.com/my-reads/cloudflare-kitesurf-agent-browser/Kitesurf is interesting because it asks what a browser should look like when the primary user is an agent, not a human.Thu, 03 Sep 2026 18:30:00 GMTCloudflareBrowserAgentsWebhttps://soumitrasaha.com/my-reads/cloudflare-workers-ai-kimi-k25/https://soumitrasaha.com/my-reads/cloudflare-workers-ai-kimi-k25/Cloudflare's large-model Workers AI post is useful because it connects model choice, inference cost, and agent infrastructure in one practical story.Thu, 03 Sep 2026 18:30:00 GMTCloudflareWorkers AIKimiAgentshttps://soumitrasaha.com/my-reads/deepseek-v4-preview-1m-context-open-model-race/https://soumitrasaha.com/my-reads/deepseek-v4-preview-1m-context-open-model-race/DeepSeek's V4 Preview is another signal that Chinese AI labs are competing hard on context length, active parameters, and cost.Thu, 03 Sep 2026 18:30:00 GMTDeepSeekChinese AI LabsOpen ModelsLong Contexthttps://soumitrasaha.com/my-reads/google-gemini-3-api-updates-for-agents/https://soumitrasaha.com/my-reads/google-gemini-3-api-updates-for-agents/A useful Google Developers read on thinking control, media resolution, thought signatures, and structured output with grounding.Thu, 03 Sep 2026 18:30:00 GMTGoogleGeminiAPIAgentshttps://soumitrasaha.com/my-reads/google-gemini-embedding-2-multimodal-rag/https://soumitrasaha.com/my-reads/google-gemini-embedding-2-multimodal-rag/A Google Developers post about one embedding model for text, images, video, audio, and documents, which is very relevant for agentic RAG.Thu, 03 Sep 2026 18:30:00 GMTGoogleGeminiEmbeddingsRAGhttps://soumitrasaha.com/my-reads/google-gemini-nano-frozen-mtp/https://soumitrasaha.com/my-reads/google-gemini-nano-frozen-mtp/A useful on-device AI read about accelerating frozen production models with multi-token prediction instead of training a separate drafter.Thu, 03 Sep 2026 18:30:00 GMTGoogleOn-device AIGemini NanoInferencehttps://soumitrasaha.com/my-reads/google-research-io-2026/https://soumitrasaha.com/my-reads/google-research-io-2026/A catch-up read on Google's research direction across Gemini for Science, computational discovery, Earth AI, and useful applied AI systems.Thu, 03 Sep 2026 18:30:00 GMTGoogle ResearchGeminiScienceAIhttps://soumitrasaha.com/my-reads/huggingface-state-of-open-models-summer-2026/https://soumitrasaha.com/my-reads/huggingface-state-of-open-models-summer-2026/A strong open-model ecosystem read about adoption, derivatives, small models, agent users, and why Qwen has become so important.Thu, 03 Sep 2026 18:30:00 GMTHugging FaceOpen ModelsAI EcosystemQwenhttps://soumitrasaha.com/my-reads/minimax-m25-cheap-long-running-agent-work/https://soumitrasaha.com/my-reads/minimax-m25-cheap-long-running-agent-work/MiniMax M2.5 is interesting because it pushes the cost side of agentic coding and productivity work very aggressively.Thu, 03 Sep 2026 18:30:00 GMTMiniMaxChinese AI LabsAgentsCosthttps://soumitrasaha.com/my-reads/openai-codex-agent-loop-deep-dive/https://soumitrasaha.com/my-reads/openai-codex-agent-loop-deep-dive/This OpenAI post is a useful technical read because it explains the agent loop, context construction, model calls, and tool orchestration behind Codex.Thu, 03 Sep 2026 18:30:00 GMTOpenAICodexAgentsDeveloper Toolshttps://soumitrasaha.com/my-reads/openai-gpt-56-builders-guide/https://soumitrasaha.com/my-reads/openai-gpt-56-builders-guide/A practical read on model selection, reasoning controls, multi-agent orchestration, and cost-aware AI product design.Thu, 03 Sep 2026 18:30:00 GMTOpenAIGPTAgentsModel Routinghttps://soumitrasaha.com/my-reads/openai-gpt-oss-open-weight-models/https://soumitrasaha.com/my-reads/openai-gpt-oss-open-weight-models/OpenAI's gpt-oss release is worth tracking because it puts strong reasoning models into open-weight deployment paths with an Apache 2.0 license.Thu, 03 Sep 2026 18:30:00 GMTOpenAIOpen WeightsReasoningLocal AIhttps://soumitrasaha.com/my-reads/openai-new-tools-for-building-agents/https://soumitrasaha.com/my-reads/openai-new-tools-for-building-agents/This OpenAI article is useful as a baseline for Responses API, built-in tools, tracing, evaluations, and how agent platforms are getting structured.Thu, 03 Sep 2026 18:30:00 GMTOpenAIAgentsResponses APIEvaluationhttps://soumitrasaha.com/my-reads/qwen38-max-coding-cowork/https://soumitrasaha.com/my-reads/qwen38-max-coding-cowork/Qwen's newer releases are worth following because the ecosystem around Qwen models is becoming a default base for many open-model builders.Thu, 03 Sep 2026 18:30:00 GMTQwenChinese AI LabsCoding ModelsOpen Modelshttps://soumitrasaha.com/my-reads/swebench-aider-coding-model-leaderboards/https://soumitrasaha.com/my-reads/swebench-aider-coding-model-leaderboards/SWE-bench and Aider are both useful for coding model comparison because they test patching, editing, and multi-language problem solving in different ways.Thu, 03 Sep 2026 18:30:00 GMTModel ComparisonsCodingSWE-benchAiderhttps://soumitrasaha.com/my-reads/zai-glm-53-frontier-coding/https://soumitrasaha.com/my-reads/zai-glm-53-frontier-coding/GLM-5.3 is useful to track because Z.ai is competing directly in coding, agentic engineering, and efficient open-weight models.Thu, 03 Sep 2026 18:30:00 GMTZ.aiGLMChinese AI LabsCoding Modelshttps://soumitrasaha.com/blogs/building-backends-with-bun-and-hono-without-making-them-heavy/https://soumitrasaha.com/blogs/building-backends-with-bun-and-hono-without-making-them-heavy/A practical way I like to structure Bun and Hono backends: small routes, typed validation, clear errors, and no unnecessary framework weight.Wed, 02 Sep 2026 18:30:00 GMTBunHonoBackendTypeScripthttps://soumitrasaha.com/blogs/what-i-learned-building-a-production-react-native-fintech-app/https://soumitrasaha.com/blogs/what-i-learned-building-a-production-react-native-fintech-app/Notes from building and owning a real React Native fintech app: payments, native modules, releases, observability, and the small things that keep users safe.Tue, 01 Sep 2026 18:30:00 GMTReact NativeFintechMobilePaymentshttps://soumitrasaha.com/blogs/rust-ropes-and-dirty-regions-terminal-ui-engine/https://soumitrasaha.com/blogs/rust-ropes-and-dirty-regions-terminal-ui-engine/A practical look at the rendering ideas behind a fast terminal UI engine: layout, framebuffers, dirty regions, text editing, and why Rust fits the core.Mon, 31 Aug 2026 18:30:00 GMTRustTerminal UIPerformanceBetterTUIhttps://soumitrasaha.com/blogs/why-i-like-static-first-websites-in-2026/https://soumitrasaha.com/blogs/why-i-like-static-first-websites-in-2026/Static first websites are still one of the best choices for portfolios, blogs, docs, and content-heavy products because they are fast, simple, and easy to trust.Sun, 30 Aug 2026 18:30:00 GMTAstroSEOPerformanceFrontendhttps://soumitrasaha.com/blogs/agentic-coding-needs-harness-engineering-not-just-better-prompts/https://soumitrasaha.com/blogs/agentic-coding-needs-harness-engineering-not-just-better-prompts/My practical view on coding agents: prompts help, but the real quality comes from tools, tests, context, permissions, and a good harness loop.Sat, 29 Aug 2026 18:30:00 GMTAI AgentsDeveloper ToolsTestingArchitecturehttps://soumitrasaha.com/blogs/designing-a-database-layer-that-survives-product-changes/https://soumitrasaha.com/blogs/designing-a-database-layer-that-survives-product-changes/The database layer should keep product code stable when schema, queries, and business rules change. This is how I like to think about it.Fri, 28 Aug 2026 18:30:00 GMTDatabaseBackendArchitectureTypeScripthttps://soumitrasaha.com/blogs/how-i-evaluate-ai-models-for-coding-work/https://soumitrasaha.com/blogs/how-i-evaluate-ai-models-for-coding-work/Benchmarks are useful, but I judge coding models by how they behave inside real repositories: reading context, editing safely, testing properly, and admitting uncertainty.Thu, 27 Aug 2026 18:30:00 GMTAI ModelsCoding AgentsBenchmarksDeveloper Toolshttps://soumitrasaha.com/blogs/from-feature-branch-to-production-my-shipping-checklist/https://soumitrasaha.com/blogs/from-feature-branch-to-production-my-shipping-checklist/The checklist I like before shipping: branch discipline, scoped changes, tests, lint, typecheck, build, smoke testing, and honest release notes.Wed, 26 Aug 2026 18:30:00 GMTShippingTestingGitEngineeringhttps://soumitrasaha.com/blogs/the-kind-of-projects-i-want-to-build-next/https://soumitrasaha.com/blogs/the-kind-of-projects-i-want-to-build-next/The work I am most interested in now: AI-native products, serious frontend systems, backend platforms, Rust developer tools, and products with real users.Tue, 25 Aug 2026 18:30:00 GMTFreelanceAI NativeRustFull Stack