Senior Engineer turned Agentic AI Architect — I build Voice AI Agents, Multi-Agent Systems, and Developer Tools. With 30+ hackathon wins, I love turning bold ideas into real products at the intersection of AI and engineering.https://bvsbharat.com/https://bvsbharat.com/posts/2026/what-survives-compaction/https://bvsbharat.com/posts/2026/what-survives-compaction/June's research reframes context management: the discard step is now where both agent quality and safety quietly leak.Wed, 01 Jul 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/the-cheapest-agent-upgrade-is-a-stop-condition/https://bvsbharat.com/posts/2026/the-cheapest-agent-upgrade-is-a-stop-condition/Mid-2026 data keeps pointing the same way: bounding an agent's loop beats unleashing it. Turn limits and budgets buy more than a bigger model.Tue, 30 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/ten-agents-three-merges/https://bvsbharat.com/posts/2026/ten-agents-three-merges/This month's agent tools made spawning parallel coding agents trivial. The constraint moved to the merge decision—and that doesn't parallelize.Mon, 29 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-security-moved-to-the-action-layer/https://bvsbharat.com/posts/2026/agent-security-moved-to-the-action-layer/Runtime authorization — intercepting tool calls before they execute — is becoming the real security boundary for agents, and a standard is forming fast.Sun, 28 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/computer-use-agents-crossed-human-parity/https://bvsbharat.com/posts/2026/computer-use-agents-crossed-human-parity/Frontier models now beat the human baseline on OSWorld-Verified — but the benchmark just got rebuilt, and the architecture quietly shifted off pixels.Sat, 27 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-catches-others-mistakes-not-its-own/https://bvsbharat.com/posts/2026/agent-catches-others-mistakes-not-its-own/New research says self-correction fails because of the role label on the claim, not the claim's content. The fix is structural, and cheaper than you think.Fri, 26 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-benchmark-score-is-an-experiment/https://bvsbharat.com/posts/2026/agent-benchmark-score-is-an-experiment/Recent work shows a single agent leaderboard number is wrong three independent ways: it's noisy, it's overfit, and the judge measuring it is unreliable.Thu, 25 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-learned-memory-policy/https://bvsbharat.com/posts/2026/agent-learned-memory-policy/A June 2026 wave moves the store/evict/retrieve decision from heuristics to a trained policy, and pushes consolidation into an offline sleep phase.Sun, 21 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/injection-stopped-being-single-turn/https://bvsbharat.com/posts/2026/injection-stopped-being-single-turn/Once agents got long-term memory, a one-time prompt injection could survive across sessions. Mid-2026 research shows both the attack and the defense moving up the stack.Sat, 20 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/the-agent-stopped-waiting-to-be-asked/https://bvsbharat.com/posts/2026/the-agent-stopped-waiting-to-be-asked/June 2026 mainstreamed always-on agents that listen to event streams instead of prompts — and that one change breaks the trigger, trust, and latency models all at once.Fri, 19 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/context-window-grew-a-memory-manager/https://bvsbharat.com/posts/2026/context-window-grew-a-memory-manager/A June 2026 wave of papers shows pruning beats full context on accuracy and cost — and that eviction is becoming a deterministic, cache-aware system, not a summarize call.Thu, 18 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/skill-supply-chain-got-poisoned-first/https://bvsbharat.com/posts/2026/skill-supply-chain-got-poisoned-first/Agent skills are an executable supply chain that runs with your agent's full privileges — and the first wave of benchmarks shows our defenses see only half the attacks.Wed, 17 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/optimization-moving-from-weights-to-english/https://bvsbharat.com/posts/2026/optimization-moving-from-weights-to-english/Recent work turns skills, harnesses, and context into objects you can search over and benchmark — optimizing the English around a frozen model instead of the model.Tue, 16 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/code-is-the-action-space-now/https://bvsbharat.com/posts/2026/code-is-the-action-space-now/Frameworks are quietly replacing JSON tool calls with generated code. That collapses turns and tokens — and pushes isolation down to the single call.Mon, 15 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/the-harness-got-a-name/https://bvsbharat.com/posts/2026/the-harness-got-a-name/A new survey and Microsoft's BUILD 2026 release both landed on the same idea: agent capability is leaving the model and moving into the harness.Mon, 15 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/skills-are-the-new-sdk/https://bvsbharat.com/posts/2026/skills-are-the-new-sdk/OpenAI killed its visual Agent Builder the same week Google shipped first-party skills. The agent capability layer just consolidated.Fri, 12 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-doesnt-know-when-its-failing/https://bvsbharat.com/posts/2026/agent-doesnt-know-when-its-failing/New benchmarks measure calibrated refusal and premature self-stops, and the data says agent confidence signals are broken. Here's how to engineer around it.Thu, 11 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/the-agent-got-its-own-account/https://bvsbharat.com/posts/2026/the-agent-got-its-own-account/In ten days of June 2026, agents got their own budget, their own permission manifest, and their own credentials. The agent is now a principal, not a feature.Wed, 10 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-rl-environment-synthesis/https://bvsbharat.com/posts/2026/agent-rl-environment-synthesis/Agent RL's bottleneck moved from data to reward to the environment itself. The newest research tries to take humans out of environment-building entirely.Tue, 09 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/data-agents-discovery-not-sql/https://bvsbharat.com/posts/2026/data-agents-discovery-not-sql/Anthropic and OpenAI independently shipped internal data agents and reached the same conclusion: discovery beats generation, and structure beats access.Mon, 08 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-trajectory-cost-center/https://bvsbharat.com/posts/2026/agent-trajectory-cost-center/A wave of mid-2026 research stopped trying to make the model cheaper and started compressing the agent's own trajectory — at the observation, action, and skill level.Sat, 06 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/compiling-the-agent-loop-away/https://bvsbharat.com/posts/2026/compiling-the-agent-loop-away/Three late-May 2026 papers attack the agent loop itself — compiling it into weights, speculating through idle time, and letting agents rewrite their own source.Fri, 05 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-parallelism-orchestration-as-code/https://bvsbharat.com/posts/2026/agent-parallelism-orchestration-as-code/Late May 2026 made parallel fan-out the agent's main scaling axis — orchestration moved into code, tests became the gate, and the meter started running.Thu, 04 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-stack-state-shells-shortcuts/https://bvsbharat.com/posts/2026/agent-stack-state-shells-shortcuts/MCP went stateless, a wave of coding-agent RCEs landed, and a new benchmark measured reward hacking — the three properties that make an agent useful all became liabilities.Wed, 03 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agents-write-their-own-skills/https://bvsbharat.com/posts/2026/agents-write-their-own-skills/May 2026's skill-library research shows agents can now accumulate reusable capabilities, but retrieving and adopting them is harder than generating them.Tue, 02 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/where-the-agent-loop-runs/https://bvsbharat.com/posts/2026/where-the-agent-loop-runs/The week of May 19 separated the agent loop from tool execution. Whoever hosts the loop now owns your latency, reliability, and lock-in.Mon, 01 Jun 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/capability-reliability-gap/https://bvsbharat.com/posts/2026/capability-reliability-gap/New work argues agents are measured wrong: accuracy keeps climbing while consistency, robustness, and predictability barely move. The fix is architectural.Sun, 31 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/verification-is-the-agent-substrate/https://bvsbharat.com/posts/2026/verification-is-the-agent-substrate/The agents scaling fastest in mid-2026 share one trait: their output lands in a column a machine can check. The verifier, not the model, is the moat.Sat, 30 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-rl-reward-design-split/https://bvsbharat.com/posts/2026/agent-rl-reward-design-split/Recent papers disagree on whether to reward agents per-turn or only at the end — and the answer reveals where RL for agents is actually headed.Fri, 29 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-benchmark-reckoning/https://bvsbharat.com/posts/2026/agent-benchmark-reckoning/STATE-Bench, DeepSWE, Agent Island, SWE-bench Live: a wave of new evals exposes how much the old leaderboards were inflating.Thu, 28 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-is-a-workload-not-a-script/https://bvsbharat.com/posts/2026/agent-is-a-workload-not-a-script/Mid-May 2026 quietly shipped the operations layer for agents — versioned environments, runtime drain, behavior-based evals, portable skills.Wed, 27 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agent-trust-stack-materialized/https://bvsbharat.com/posts/2026/agent-trust-stack-materialized/Skill cards, self-hosted sandboxes, MCP tunnels, computer-use verifiers, and a Five Eyes warning all landed in twenty-one days. The boring perimeter around capable agents finally has shape.Tue, 26 May 2026 18:00:00 GMThttps://bvsbharat.com/posts/2026/agentic-web-end-of-browsing/https://bvsbharat.com/posts/2026/agentic-web-end-of-browsing/Twenty minutes of tabs vs. five minutes of prompt. The traditional web wasn't designed for humans — it was designed for mice. The agent-native web is quietly dismantling the parts that never made sense.Tue, 26 May 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/21-future-of-agent-architectures/https://bvsbharat.com/posts/2026/agents-arch/21-future-of-agent-architectures/Five months of weekly posts; five trends worth watching. What 2026 settled, what it didn't, and the architectural bets I'd make if I were starting now.Mon, 25 May 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/long-horizon-agents-tasks-that-run-for-hours/https://bvsbharat.com/posts/2026/long-horizon-agents-tasks-that-run-for-hours/Six-hour agent runs are now real. The harness — checkpoints, durable state, recovery — matters more than the model. A field guide to the long-running pattern.Thu, 21 May 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/20-reference-architecture/https://bvsbharat.com/posts/2026/agents-arch/20-reference-architecture/Eighteen weeks of patterns combined into one production-shaped architecture. The full diagram, the repo layout, the contracts between components, and the migration sequence.Mon, 18 May 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/19-framework-comparison/https://bvsbharat.com/posts/2026/agents-arch/19-framework-comparison/Four agent frameworks, four different bets. The honest comparison — what each gets right, where each leaks, and the decision tree for picking one (or composing several).Mon, 11 May 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/skills-connectors-subagents-template/https://bvsbharat.com/posts/2026/skills-connectors-subagents-template/Anthropic just shipped 10 financial services agent templates. The interesting part isn't the templates — it's the three-layer architecture quietly becoming the standard for enterprise agents.Mon, 11 May 2026 07:43:18 GMThttps://bvsbharat.com/posts/2026/code-with-claude-2026-recap/https://bvsbharat.com/posts/2026/code-with-claude-2026-recap/Anthropic shipped a lot on May 6 — Managed Agents updates, Dreaming, Outcomes, Multi-agent Orchestration, and a SpaceX partnership. The signal-to-noise filtered down to five things that change how you build.Thu, 07 May 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/18-hierarchical-supervisor-workers/https://bvsbharat.com/posts/2026/agents-arch/18-hierarchical-supervisor-workers/When one agent isn't enough, hierarchy is usually the first answer. Supervisor agents, worker pools, and the patterns that keep multi-agent systems coherent at scale.Mon, 04 May 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agent-observability-tracing-replay/https://bvsbharat.com/posts/2026/agent-observability-tracing-replay/Langfuse got acquired by ClickHouse. Helicone hit maintenance mode. OpenTelemetry standardized LLM tracing. The observability stack for agents reshuffled in three months. Here's what it looks like now.Fri, 01 May 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/17-enterprise-governance/https://bvsbharat.com/posts/2026/agents-arch/17-enterprise-governance/Who decides what agents are allowed to do, how those decisions get enforced at runtime, and how to prove it later. The 2026 governance stack — policy as code, audit trails, and the EU AI Act bit.Mon, 27 Apr 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agent-evals-beyond-llm-as-judge/https://bvsbharat.com/posts/2026/agent-evals-beyond-llm-as-judge/Vibes-based scoring is finally dying. Trajectory eval, rubric eval, golden replay, and the test pyramid that production agent teams actually run.Fri, 24 Apr 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/16-security-for-agents/https://bvsbharat.com/posts/2026/agents-arch/16-security-for-agents/Agents are systems that take instructions from untrusted text and execute actions. The 2026 threat model — prompt injection that's not solved, tool sandboxing, scoped credentials, and authorization that survives contact with the LLM.Mon, 20 Apr 2026 16:00:00 GMThttps://bvsbharat.com/posts/cascaded-vs-fused-voice-agents-2026/https://bvsbharat.com/posts/cascaded-vs-fused-voice-agents-2026/Deep dive into voice agent architectures. Why cascaded models give you control and fused models trade complexity for naturalness. What we're learning from shipping production agents at scale.Fri, 17 Apr 2026 21:16:05 GMThttps://bvsbharat.com/posts/2026/agents-arch/15-cost-latency-engineering/https://bvsbharat.com/posts/2026/agents-arch/15-cost-latency-engineering/Agents bill non-linearly. The patterns that matter — prompt caching, tiered model routing, parallel tool calls, retrieval budgets — and the dashboards that catch waste before it ships.Mon, 13 Apr 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/sandbox-execution-grew-up/https://bvsbharat.com/posts/2026/sandbox-execution-grew-up/Firecracker microVMs, gVisor containers, persistent workspaces, and the $24M Series A nobody quite expected. The sandbox layer beneath every serious agent — and how to pick the right one.Fri, 10 Apr 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/14-deployment-patterns/https://bvsbharat.com/posts/2026/agents-arch/14-deployment-patterns/Three deployment shapes for production agents — request/response, long-running stateful, and background batch — with the trade-offs around Cloud Run, Kubernetes, managed runtimes, and queue workers.Mon, 06 Apr 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/prompting-voice-agents-realistic-speech/https://bvsbharat.com/posts/2026/prompting-voice-agents-realistic-speech/Why your cascaded voice agent sounds robotic — and how to fix it with concrete examples, SSML pause patterns, emotion tags, and personality-as-behavior prompting techniques.Fri, 03 Apr 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/13-evaluating-agents/https://bvsbharat.com/posts/2026/agents-arch/13-evaluating-agents/You can't ship agents you can't measure. The 2026 eval stack — task-level scoring, trajectory grading, LLM-as-judge with calibration, and the regression gates that catch silent quality drops.Mon, 30 Mar 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/cost-optimized-agent-architectures/https://bvsbharat.com/posts/2026/cost-optimized-agent-architectures/Caching, routing, distillation, and per-task model selection. The four moves that take a $0.40/task agent to $0.04/task without anyone noticing the difference.Thu, 26 Mar 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/12-agent-observability/https://bvsbharat.com/posts/2026/agents-arch/12-agent-observability/Agents are non-deterministic stateful systems calling external services. You cannot debug them without tracing. The 2026 observability stack — LangSmith, Langfuse, OpenTelemetry — and the metrics that matter.Mon, 23 Mar 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/web-research-agents-state-of-the-art/https://bvsbharat.com/posts/2026/web-research-agents-state-of-the-art/Operator died, Browser Use became the default substrate, Manus shipped at scale, and the gap between demo and reliable production narrowed considerably. A field report.Thu, 19 Mar 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/11-rag-for-agents/https://bvsbharat.com/posts/2026/agents-arch/11-rag-for-agents/RAG for agents is not RAG for chatbots. Hybrid retrieval, query rewriting, retrieval-as-tool, and the patterns that beat 'top-k from a vector store' in 2026.Mon, 16 Mar 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/deep-agents-planner-executor-critic/https://bvsbharat.com/posts/2026/deep-agents-planner-executor-critic/The three-role pattern that powered Manus, then LangChain Deep Agents, then half the production agents shipping in early 2026. Why it works, when it doesn't, and how to actually build one.Thu, 12 Mar 2026 15:00:00 GMThttps://bvsbharat.com/posts/2026/context-engineering-agents/https://bvsbharat.com/posts/2026/context-engineering-agents/A deep dive into context engineering — the techniques that separate toy demos from production AI agents. Covers compaction, offloading, isolation, caching, and prioritization with real examples from Manus, Claude Code, and Devin.Wed, 25 Feb 2026 08:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/10-multi-agent-communication-a2a-mcp/https://bvsbharat.com/posts/2026/agents-arch/10-multi-agent-communication-a2a-mcp/When one agent isn't enough, how do they talk? A2A for peer-to-peer delegation, MCP for tools, and the durable message-bus patterns underneath multi-agent systems in 2026.Mon, 09 Mar 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/a-deep-dive-into-multi-agent-reinforcement-learning-with-grpo/https://bvsbharat.com/posts/2026/a-deep-dive-into-multi-agent-reinforcement-learning-with-grpo/How exploring LLM fine-tuning led to building a Gymnasium-compatible RL environment where 7 LLM-powered agents run a company — trained with GRPO + LoRA on Qwen 2.5 14B — and what we learned about reward design, emergent collaboration, and the future of agentic AI.Sat, 07 Mar 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/09-tool-use-patterns/https://bvsbharat.com/posts/2026/agents-arch/09-tool-use-patterns/Tools are what turn a chatbot into an agent. The 2026 stack for tool design — small scoped tool sets, structured arguments, MCP servers, and the failure modes you only see in production.Mon, 02 Mar 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/mcp-too-many-tools-problem/https://bvsbharat.com/posts/2026/mcp-too-many-tools-problem/AI agents are drowning in tools. The more APIs you connect via MCP, the worse your agent performs. Here's why, and what Code Mode changes.Tue, 24 Feb 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/08-agent-memory-architectures/https://bvsbharat.com/posts/2026/agents-arch/08-agent-memory-architectures/Three kinds of memory, four-tier consolidation pipelines, hybrid retrieval that beat pure vector search, and the unsolved staleness problem. The 2026 state of the art for agents that learn over time.Mon, 23 Feb 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/ai-app-paradox-experience/https://bvsbharat.com/posts/2026/ai-app-paradox-experience/We've been so obsessed with what AI can do that we forgot about how it feels to use it. The AI experience layer is the next frontier — not the model, not the capabilities.Fri, 20 Feb 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/07-managed-agents-bedrock-claude/https://bvsbharat.com/posts/2026/agents-arch/07-managed-agents-bedrock-claude/Claude Managed Agents and Amazon Bedrock AgentCore turned 'run an agent in production' from a months-long platform project into a credit-card decision. Here's how they differ and when each wins.Mon, 16 Feb 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/tool-selection-at-scale-200-tools/https://bvsbharat.com/posts/2026/tool-selection-at-scale-200-tools/Past ~30 tools, agent reliability falls off a cliff. Past ~100, it's chaos. Here's the actual engineering — RAG-over-tools, semantic routing, dynamic loading, and namespacing — that production teams ship to stay sane.Thu, 12 Feb 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/06-google-adk-architecture/https://bvsbharat.com/posts/2026/agents-arch/06-google-adk-architecture/Google's Agent Development Kit isn't another agent loop. It's an event-driven runtime, a multi-language SDK, and a deploy target that lines up with Cloud Run and GKE. Here's how the pieces fit.Mon, 09 Feb 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/sub-agents-new-microservices/https://bvsbharat.com/posts/2026/sub-agents-new-microservices/The orchestrator-worker pattern that took over agent design in late 2025 is the same pattern that took over backend design in 2014. The wins are real. So are the failure modes.Thu, 05 Feb 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/05-crewai-role-based-orchestration/https://bvsbharat.com/posts/2026/agents-arch/05-crewai-role-based-orchestration/Crews, Flows, and the dual-layer architecture that lets CrewAI hold its own against LangGraph in 2026 — when role-based composition is the right abstraction and when it isn't.Mon, 02 Feb 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/testing-open-source-ai-agent-sdks/https://bvsbharat.com/posts/2026/testing-open-source-ai-agent-sdks/A comprehensive hands-on comparison of seven open-source AI agent frameworks — which one should you actually use?Thu, 29 Jan 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/04-langgraph-patterns-hitl/https://bvsbharat.com/posts/2026/agents-arch/04-langgraph-patterns-hitl/The three control-flow patterns that turn a toy LangGraph agent into a production system — bounded cycles, parallel fan-out/merge branches, and interrupt-resume human approval gates.Mon, 26 Jan 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/agent-framework-decision-tree-2026/https://bvsbharat.com/posts/2026/agent-framework-decision-tree-2026/Six serious frameworks, four orchestration styles, and one tired question I keep getting asked. Here's the decision tree I actually use.Thu, 22 Jan 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/03-langgraph-deep-dive/https://bvsbharat.com/posts/2026/agents-arch/03-langgraph-deep-dive/LangGraph turns the agent loop into an explicit state graph — nodes, edges, reducers, checkpointers. The mental model, the type system, and the patterns that make it production-grade.Mon, 19 Jan 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/mcp-ecosystem-crossed-inflection-point/https://bvsbharat.com/posts/2026/mcp-ecosystem-crossed-inflection-point/Fourteen months in, the Model Context Protocol stopped being a curiosity and started being plumbing. Here's what changed over the holidays — registries, governance, and the first scaling pains.Thu, 15 Jan 2026 16:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/02-langchain-in-production/https://bvsbharat.com/posts/2026/agents-arch/02-langchain-in-production/LangChain is no longer the agent framework — LangGraph is. But its primitives (runnables, retrievers, callbacks) became the connective tissue underneath. Here's what to keep and what to drop in 2026.Mon, 12 Jan 2026 17:00:00 GMThttps://bvsbharat.com/posts/2026/jarvis-iot-agentic-ai/https://bvsbharat.com/posts/2026/jarvis-iot-agentic-ai/Open-sourcing my childhood dream — an AI agent that understands context, makes decisions, and controls connected devices just like JARVIS.Sat, 10 Jan 2026 00:00:00 GMThttps://bvsbharat.com/posts/2026/agents-arch/01-anatomy-of-an-ai-agent/https://bvsbharat.com/posts/2026/agents-arch/01-anatomy-of-an-ai-agent/What actually goes inside a production AI agent — the reasoning loop, the tool layer, memory, control flow, and the surface where models meet your business systems.Mon, 05 Jan 2026 17:00:00 GMThttps://bvsbharat.com/posts/2023/optimizing-useeffect-react-deep-dive/https://bvsbharat.com/posts/2023/optimizing-useeffect-react-deep-dive/Best practices and strategies to optimize React's useEffect hook — minimizing dependencies, cleaning up side effects, custom hooks, useMemo, debouncing, and more.Mon, 28 Aug 2023 00:00:00 GMT