Essays on AI governance, measurement and data qualityhttps://readsigil.com/2026-10-13T00:00:00+05:30Shashwat Kumar https://readsigil.com/essays/brakes.html2026-10-13T00:00:00+05:30

AI governance and human-in-the-loop design: why more capable AI needs better brakes, and three questions for deciding what a system may do on its own.

Shashwat Kumar https://readsigil.com/essays/many-agents.html2026-10-13T00:00:00+05:30

Multi-agent AI systems: why intelligence may emerge from orchestration between specialised agents rather than one model, and where the risk sits.

Shashwat Kumar https://readsigil.com/essays/goodhart.html2026-10-13T00:00:00+05:30

Goodhart's law and agentic AI: how optimising systems game metrics, and why counter-metrics and a portfolio of measures are the defence.

Shashwat Kumar https://readsigil.com/essays/agents-trade.html2026-10-13T00:00:00+05:30

Agentic AI in banking and payments: when software agents negotiate and pay, how to design identity, authority limits and audit trails.

Shashwat Kumar https://readsigil.com/essays/data-quality.html2026-10-13T00:00:00+05:30

Data quality for AI and analytics: metric dictionaries, pipeline checks, data ownership and known-answer tests that make data boring and trusted.

Shashwat Kumar https://readsigil.com/essays/attention.html2026-10-13T00:00:00+05:30

The attention economy as a measurement problem: why attention metrics drift from lasting value, and what that means for careers and incentives.

Shashwat Kumar https://readsigil.com/magazine/staged-release.html2026-10-13T09:00:00+05:30

Google, Mistral and OpenAI each shipped frontier capability this month behind deliberate gates. The gate is becoming as important as the model.

Shashwat Kumar https://readsigil.com/magazine/intent-before-payment.html2026-10-13T09:00:00+05:30

American Express wants the customer's instruction captured and signed before an agent pays. India is reportedly designing agent payments on UPI with spending limits and audit trails. The plumbing of trust is being rebuilt for machine customers.

Shashwat Kumar https://readsigil.com/magazine/personal-agents-memory.html2026-10-12T09:00:00+05:30

Meta's Muse Wonder pairs 'Active Memory' with 'Active Focus'. Memory is what makes an agent useful, and it is also what makes it a data-governance problem.

Shashwat Kumar https://readsigil.com/magazine/tokens-per-megawatt.html2026-10-12T09:00:00+05:30

Microsoft is making its own Maia chips the primary inference system for its products, and framing the case in tokens per watt. The choice of metric tells you what the industry now optimises for.

Shashwat Kumar https://readsigil.com/magazine/agents-get-an-os.html2026-10-08T09:00:00+05:30

Microsoft shipped execution containers for agents on Windows, NVIDIA put a petaflop in a laptop, and Google Cloud launched a 'universal agent for work'. The agent is moving from a chat window into the plumbing.

Shashwat Kumar https://readsigil.com/magazine/open-weights-quiet-month.html2026-10-08T09:00:00+05:30

A 740-million-parameter multimodal embedding model under Apache 2.0, open frontier weights promised after red-teaming, and free AI security scans for open-source projects. Open AI infrastructure is maturing at the edges.

Shashwat Kumar https://readsigil.com/magazine/intelligent-ui.html2026-10-07T09:00:00+05:30

GPT-6 in ChatGPT now answers with buttons, forms and charts. The output of a model is turning into a small piece of software, and that changes how we should judge it.

Shashwat Kumar https://readsigil.com/magazine/price-of-intelligence.html2026-10-07T09:00:00+05:30

Anthropic's Haiku 5.5 and xAI's Grok 4.7 both lead with price and speed. Cheaper capable models change what is worth automating, and what is worth measuring.

Shashwat Kumar https://readsigil.com/magazine/indic-documents.html2026-09-24T09:00:00+05:30

Sarvam's Vision 2.1 targets the last mile of document intelligence: Indic handwriting, forms and multi-page tables. Its most useful contribution may be the open benchmark that came with it.

Shashwat Kumar https://readsigil.com/magazine/quantum-manufacturing.html2026-10-13T09:00:00+05:30

IonQ, Quantinuum, IBM's Anderon and D-Wave all made manufacturing news in five weeks. When an industry starts talking about tapeouts and wafer lots, it is preparing to be measured like one.

Shashwat Kumar https://readsigil.com/magazine/robotaxi-capital.html2026-10-08T09:00:00+05:30

Waymo raised its first debt, began driverless testing in Detroit and announced Singapore for 2028. Autonomy is being financed and rolled out like infrastructure, city by city.

Shashwat Kumar https://readsigil.com/magazine/quantum-verification.html2026-10-07T09:00:00+05:30

IBM moved to the testing stage of DARPA's benchmarking programme, and Google researchers proposed a software trick against residual logical errors. The field is shifting from claims to independent checks.

Shashwat Kumar https://readsigil.com/magazine/humanoid-hype-gap.html2026-10-07T09:00:00+05:30

Reports say Tesla is ramping Optimus production, while Google DeepMind's latest robotics model controls whole humanoid bodies. The distance between production targets and reliable work is the number to track.

Shashwat Kumar https://readsigil.com/magazine/india-quantum-km.html2026-09-30T09:00:00+05:30

The National Quantum Mission says it has crossed 1,000 km of quantum-secure communication in about three years. A free-space link test followed. The milestone is real progress; the definition of the metric matters.

Shashwat Kumar https://readsigil.com/magazine/helix-30-homes.html2026-09-17T09:00:00+05:30

Figure's Helix 2.5 took a humanoid into 30 homes it had never seen. The most impressive thing about the announcement is how it measured itself.

Shashwat Kumar