Bhargav Joshi
Full-stack developer, Geiger Studios
Bhargav Joshi is a full-stack developer working across Node.js and PHP/Laravel, and increasingly on the AI systems behind web apps: RAG pipelines, self-hosted inference and custom agents. Bhargav builds Geiger Studios, a connected suite of apps for planning, creating and running work, and writes here about forms, events, AI models and what it takes to ship them.
Articles
Conditional logic in forms: 7 examples
How conditional logic works in online forms: the parts of a rule, which rule to use when, seven examples to copy, and the bugs to test for before launch.
Model ReviewsGPT‑5.6 Sol cuts cost by two‑thirds but raises hallucination risk
GPT‑5.6 Sol delivers the same Intelligence Index score as Claude Fable 5 for $1.04 per task – about one‑third the price – but introduces a measurable rise in hallucinations.
Model ReviewsGrok 4.5: The cheapest near‑frontier model that also hallucinates twice as much
At $0.31 per Intelligence Index task, Grok 4.5 undercuts rivals, yet its hallucination rate jumps from 25% to 54%, raising safety concerns for cost‑focused adopters.
Model ReviewsMuse Spark 1.1 Review: Meta Scores 51 on the Intelligence Index - Benchmarks, Cost Per Task & Where It Still Lags
Meta's Muse Spark 1.1 scores 51 on the Artificial Analysis Intelligence Index at ~$0.26 per task — but its 4x factuality gain came from abstention, not accuracy.
Model ReviewsAutomationBench-AA results: AI agents still break business rules
A new benchmark from Zapier and Artificial Analysis tested AI agents on 657 real SaaS workflow tasks. Anthropic's models lead — but every model breaks business rules, and Finance automation is still the weak spot.
Model ReviewsClaude Sonnet 5 Review: Benchmarks, Cost Per Task & How It Compares to Opus 4.8
Claude Sonnet 5 scores 53 on the Artificial Analysis Intelligence Index and matches Opus 4.8 on agentic tasks — but at standard pricing it costs more per task than the model it's supposed to undercut. Here's what the numbers actually show.
Model ReviewsGLM-5.2 Review: Benchmarks, Cost & How It Compares to DeepSeek and MiniMax.
GLM-5.2 tops the open weights leaderboard with a score of 51 on the Intelligence Index. Here's a breakdown of its benchmarks, cost per task, and how it stacks up.