PinnedInTowards AIbyVenkat Peri·May 28I Built an App Entirely with Claude Code. This is What I LearntAcross 26 calendar days, I logged 97 active hours and processed 9.9 billion tokens of Claude Code on a single project. The output is an app…
PinnedInTowards AIbyVenkat Peri·Apr 2The Complete Architecture for Trustworthy Autonomous AgentsFour layers. Four questions. Missing any one of them is how production systems fail.A response icon2A response icon2
InTowards AIbyVenkat Peri·Jul 25The Constraint Always MovesAI made code generation free and verification cheap. Our velocity data says the next bottleneck is not in the toolchain at all.A response icon1A response icon1
InTowards AIbyVenkat Peri·Jun 23Once an AI Agent Removes Typing, Intent Becomes the BottleneckWhat 1,263 real coding prompts, one month of token bills, and a same-day full-stack build taught us about working with coding agents
InTowards AIbyVenkat Peri·Jun 18You Can’t Prompt Away Your LLM ProblemsWhat actually broke in a production LLM assistant, and what actually fixed itA response icon1A response icon1
InTowards AIbyVenkat Peri·Jun 13Green Evals, Wrong AnswersMaking LLM evaluations block merges when the model is nondeterministic, and the product is regulated
InTowards AIbyVenkat Peri·Jun 11When the Model Pushed BackSeven days, 535 prompts, and five times an AI codegen loop refused the architecture we proposed
InTowards AIbyVenkat Peri·Jun 10The Code Was the Easy PartWhy AI-native development spreads on apprenticeship timelines, not tooling timelines
Venkat Peri·May 30What to ask your agentic AI vendorSix procurement questions for financial services buyers, and what separates a polished demo from a shippable product
InTowards AIbyVenkat Peri·Apr 23Your Intent Classifier is Solving the Wrong ProblemWhy response architecture matters more than classification in regulated LLM apps