Answers with Costly Checks: Cheaper AI, DastAIvez and Finding Your Legal Fit

What happens when answers get cheaper, but checking them still takes time? Anthropic’s new Haiku pricing gives developers and businesses another reason to test AI on routine work. Google’s pause on certain bug reports highlights a related challenge: automated submissions still need evidence before someone can act on them.

In this issue of NuPOV, we explore that checking work alongside tools for understanding land records and finding legal help.

01. The World : Cheaper answers still need checking

Anthropic released Claude Haiku 5.5 on 7 October, with listed rates one-tenth Haiku 4.5’s: $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Longer prompts cost more. For routine document queries or classification, that makes it worth testing. The useful comparison is the total cost of getting a task right, including token use, retries and human review.

Google faced a different verification problem. From 1 October, it paused new product-vulnerability reports through its open-source bounty programme, citing increased automated submissions, most of them invalid. Google did not specify whether the submissions were AI-generated. For anyone using AI to investigate code, the lesson is practical: reproduce the suspected flaw and demonstrate its impact before reporting it. Maintainers need evidence they can act on.

02. The Shift in Bihar : Startup support and server space

ShuruKar launched on 10 October at Dr. APJ Abdul Kalam Science City, Patna, addressing a familiar difficulty for aspiring founders: having an idea but not knowing where to begin. YourStory’s The Bharat Project developed the nationwide initiative with DPIIT and Startup India, combining an AI copilot, founder mentorship and local networks. Mithila Stack participated alongside founders and aspiring entrepreneurs. The programme’s proposed route starts with a problem in the participant’s own district, then moves through refining the idea and working with mentors towards a first customer.

On 13 October, CtrlS inaugurated its ₹400-crore data centre in Patliputra Industrial Area, with a reported 12 MW IT load. It adds a local infrastructure option for businesses and institutions, whose usefulness will depend on connectivity, service terms and price. The accompanying ₹10,000-crore, 150 MW Bodh Gaya AI factory remains a separate proposal under a letter of intent, with no announced completion timeline.

03. Mithila Stack : Reading records, finding legal help

We recently shared our product portfolio and documentation on ongoing research with Shri Nitish Mishra, Bihar’s Minister for Information Technology and Urban Development and Housing, in Patna. The material connected our work to familiar needs in Bihar, including understanding land paperwork and finding legal help.

A land record can be available online and still leave its reader puzzled. DastAIvez helps users read uploaded documents through text extraction, Hindi and English explanations, page summaries and meanings of unfamiliar Farsi terms. Its jamabandi chatbot also retrieves Bihar Bhumi information through Hindi or English conversation using district, village and khata or khesra details. Voice support remains planned.

Finding an advocate brings a different uncertainty: a recommendation may give you a name without much context. Advocate Finder is designed to help people compare recorded matters and case types from official sources, alongside contact details, before deciding whom to approach. Those records offer useful context, but cannot establish a complete career history or predict a legal outcome.

04. The Hero of the Month : Earning trust before changing crops

DeHaat co-founder and CEO Shashank Kumar recalls a farmer calling before dawn, unsure whether to harvest baby corn because it still looked too small. The team had explained the crop and promised to buy it. The farmer still needed reassurance.

In a 2022 interview, Kumar describes the lesson behind such encounters: improve farmers’ returns from familiar crops before asking them to try unfamiliar ones. Their hesitation reflected the risk they carried. DeHaat also developed advisory content before investing substantially in technology.

The revealing decision was to change the sequence of the offering. A technically sound proposal became more useful when the team understood what customers could afford to risk.

What has a customer or user made you rethink recently? We’d like to hear about the change you made, and what happened next.