Viewing profile — kundan_s__r

That framing resonates a lot. In production, creativity is often just unbounded variance. Once each step is intentionally boring and constrained, failures become predictable and de…

That’s a very real example of the core problem: LLMs don’t reliably honor constraints, even when they’re explicit and simple. Instruction drift shows up fast in learning tasks — an…

A helpful way to learn this is to separate models, machines, and practice. For computation models, the circuit model and measurement-based computation cover most real work. Aaronso…

The real risk with LLMs isn’t when they fail loudly — it’s when they fail quietly and confidently, especially for non-experts or downstream systems that assume structured output eq…

That’s fair — if you’re already skeptical and paying attention, the failures are obvious and often funny. The risk tends to show up more with non-experts or downstream systems that…

This matches our experience too. The biggest reduction in hallucinations usually comes from shrinking the action space, not improving the prompt. When inputs, tools, and outputs ar…

Fair enough. A healthy dose of skepticism has served us well for every overhyped wave so far. The difference this time seems to be that AI systems don’t just fail noisily — they fa…

Really impressive work, especially on mobile. The mmap + zero-copy, read-only approach feels like the right tradeoff for files at this scale. Curious how it behaves with extremely …

This matches what I’ve seen as well. A lot of “debt relief” and “settlement” services are essentially rent-seeking intermediaries that leave consumers worse off or stuck in long pr…

That’s a very sane stance. Treating LLM output as untrusted input is probably the correct default when correctness matters. The worst failures I’ve seen happen when teams half-trus…

This is a very pragmatic take. The “90% accuracy is a liability” line resonates — in high-stakes systems, partial automation often costs more than it saves. What I like here is the…

Hi HN, For those running LLMs in real production environments (especially agentic or tool-using systems): what’s actually worked for you to prevent confident but incorrect outputs?…

This resonates. A lot of AI reading tools optimize for removal of effort (summaries, shortcuts), which often ends up weakening comprehension rather than strengthening it. One thing…

I’ve been working on Verdic Guard, a validation layer for production LLM systems where prompts, filters, and monitoring aren’t enough. In many real deployments (fintech, enterprise…

Whether or not Hallucination “happens often” depends heavily on the task domain and how you define correctness. In a simple conversational question about general knowledge, an LLM …

Interesting reflection — but I’d push back on treating surface similarities between human conversational quirks and LLM failure modes as evidence they’re really the same thing. The…

I’m building Verdic Guard to explore a problem I kept seeing with LLMs in production. Models often behave well in demos and short interactions, but once they’re embedded into long,…

I’m building Verdic Guard, an experiment around a problem I kept running into while working with LLMs in production. LLMs usually behave well in demos and short interactions, but o…

This framing resonates a lot. The core issue you’re pointing at isn’t model accuracy, it’s epistemic accountability. In most current deployments, an AI system’s output is treated a…

please check verdic.dev

We’ve been working on production LLM systems and noticed a recurring issue: even well-crafted prompts fail under real-world conditions. We wrote a technical breakdown of the failur…

We built Verdic (https://www.verdic.dev/ )after repeatedly running into the same issue while deploying LLMs in production: most AI failures aren’t about content safety, they’re abo…