Kash

Jev from TypeSafe AI answers yes/no, pick-one and rate-on-a-scale questions in 70 to 500ms for $0.042 per million input tokens. What it is, how the API works, and where it fits in an agent stack that already went single-agent.

A 136-entry bug ledger with no ticket system: what every entry carries, the eight buckets a census found, and the four-verdict triage that keeps it honest.

A production supervisor-and-specialists build: routing decisions, per-agent tool registries, 8,000 tokens down to 800, fallback models for rate limits, and streaming the agent's reasoning to the user.

No review, self-critique, supervisor review, human approval and double review, with the latency and token cost of each, and a decision matrix that picks by stakes and speed.

A zero-LLM heuristic classifier that picks an execution pattern in under 5ms, why asyncio.gather beats sub-agents until it does not, and when to expose pattern control to users.

Three execution patterns, chosen per query rather than per system: unified sequential for 80%, unified with parallel tools for 15%, multi-agent parallel for 5%. Plus why skills are orthogonal to all three.

A routing LLM call added 300ms and doubled token overhead on every request across 10,000 production queries. Why skills beat specialists for sequential work, and what removing the router changed.

Define agents as data so the directory and API stay in step, then trace every run. What per-trace token and cost attribution makes visible, and why it comes before optimisation.

A zero-tool supervisor routing to specialised agents cut tokens per request from 8,000 to 800 and took throughput from 2 requests a minute to 12. What the pattern is, and what we later learned it costs.

Multi-agent burns about 15x the tokens by design and carries 14 catalogued failure modes. Why one agent with dynamic skills, curated context and tiered memory is the 2026 default, and the read-write test for the exception.

Pydantic AI tool calls inside a LangGraph node do not emit to astream_events. Capture state from the node's return value via on_chain_end instead, in about ten lines.

A field report on a single deep-research run: where 106 agents and $76 went, why 61% of the bill was cache writes, and the one thing adversarial verification bought that a $1 query cannot.

Follow-up questions failed at nearly three times the rate of first questions in our RAG system. The cause was context split across agents, and the fix was one agent holding the whole conversation.

Tool definitions cost tokens and add decisions. Published reports put the inflection near 15 tools. The minimal-base-plus-dynamic-loading pattern, and what it changed for us.

Sequential tool calls make an agent wait on I/O it could overlap. How asyncio.gather cuts a three-lookup response from 1.5s to 0.5s, with rate limiting that keeps it safe.

How to build the data-processor list Meta's App Review asks for, why your AI vendors belong on it, and how to execute the OpenAI DPA. Anthropic's is already in its commercial terms.

A scene-by-scene demo video script for Google OAuth app verification: the two-flows rule, the fresh-account consent trap, the unverified-app warning, and the submission form.

A first-hand account of enabling Anthropic prompt caching: the trace that found the bottleneck, the three settings that fixed it, and the before-and-after cost and latency numbers.

Google reviews OAuth apps on two tracks: automated brand verification and human scope verification. What each checks, what each rejects, and the right order.

A first-hand account of getting Google Ads API Basic Access approved in two days: the application choices, the three developer-token errors, and the API version trap.

How to pass Meta business verification: why Corporation means publicly listed, where your brand name goes, which incorporation document to upload, and the domain rule for the confirmation email.

The full Meta App Review process for apps accessing business data: business verification, Tech Provider access verification, Data Handling Questions, allowed usage, screencasts, and the reviewer test account.

How to pass Meta access verification: the Tech Provider definition, the business-type multi-select, the plain-language Platform Data answer, the portfolios question, and the website rule.

Google approved our OAuth app verification in a day. Step-by-step: scope list, privacy policy, branding, justifications, demo video, submission, what changes after.