agentgateway-lint is a static linter for agentgateway config files. It reads the YAML/JSON once, runs a set of security and hygiene rules, and outputs an A-F grade. Install: git clone then python3 -m agentgateway_lint samples/risky.yaml, no dependencies needed. It addresses the gap that agentgateway itself only checks whether a config is well-formed, not whether it's dangerous.
Today's Paper Reading rotation works through arXiv:2609.11028, BenchShield — a reward-integrity detection system for LLM-agent evaluation infrastructure. It uses static, phase-aware taint analysis to expose exploitable paths before a run, paired with runtime infrastructure-side evidence to turn "this score wasn't gamed" into a verifiable claim, lifting full-chain recall from 23-94% to 77-100% on 456 human-adjudicated trajectories, with 96% runtime detection accuracy. Core concepts cover why reward hacking is an evaluation-infrastructure problem (not just a model problem), how static and dynamic taint analysis split the work, why an "optional shortcut" that breaks no stated rule is still worth defending against, and how to read a paper's numbers by asking what evidence actually backs each claim. Paired with the companion paper BAITBENCH and the real-world Hugging Face agent-swarm security incident to make the case that reward hacking isn't just an academic concern.
Temporal v1.32.0 highlights: (1) Standalone Activities reach GA, with delayed starts, operator APIs (pause/resume/reset), and batch operations; (2) Nexus callbacks now route by URL scheme by default, removing the old header-based config — a breaking, security-driven change; (3) the Unified Query Converter becomes the default, tightening type validation and empty-string filtering on Visibility queries.
Mistral closed a €3B Series D led by Samsung Electronics, with valuation jumping from €11.7B a year ago to over €21B. This round signals that Mistral isn't betting on model benchmarks — it's selling 'sovereignty' itself as a product to European governments and enterprises unwilling to hand AI workloads to US clouds.
DeepSeek-V4.1-Flash (API model: deepseek-flash): released 2026-09-10, 552B-parameter MoE with a Causal Encoder-Decoder architecture (only 8B active parameters for input, 16B for output), 1M-token context, MIT-licensed. Peak pricing: input $0.30 / output $1.20 per 1M tokens (off-peak halved) — cheaper than the predecessor V4-Flash. Terminal-Bench 4.0 jumps from 7.0 to 31.2, DeepSWE v1.1 (74.2) ties Claude Opus 5 (74.0). KV cache compressed to 890 bytes/token (1/4 of the predecessor, 437x less than DeepSeek-V1). DeepSeek announced that starting Sept 14, all traffic to its own V4-Pro flagship will be routed to this Flash model and billed at Flash rates.
The easiest way to fumble a Technical PM question is to frame technical judgment as either "engineering decides" or "I decide alone," without naming how you open the decision process to the team without losing direction. Today we use the RFC process to converge input from multiple stakeholders, and Architecture Decision Records to preserve the reasoning behind a decision, then practice an API design question about a partner system that auto-retries after a timeout and ends up creating duplicate orders — how do you make the retry safe, and how do you know the fix actually worked. The case study is Stripe's Idempotency-Key header: the same problem, solved on their payments API for a decade now, and still the reference answer people reach for on this exact question.
mcp-bi is zavora-ai's open-source Business Intelligence MCP Server. Set BI_BACKEND to switch between Superset, Metabase, Power BI, Tableau, Looker, Qlik, and QuickSight while an agent calls the same tool set across all of them. Install: cargo run (ships with a seeded fixture, no setup needed). It addresses the problem of agents only seeing a dashboard screenshot and guessing at trends.
Today's Coding rotation works through a real xAI interview question — implement dynamic batching for token decoding, refilling finished slots in a fixed-capacity batch B while correctly handling three stop conditions (stop_token, stop_sequence, max_tokens) and keeping the slot_id-to-request_id mapping correct. Core concepts cover how continuous batching trades off GPU utilization against latency, using NumPy broadcasting and boolean masks instead of explicit loops, padding and attention masks for variable-length inputs, and why a production batch inference API should scale on queue depth rather than raw GPU utilization. Bonus: a meta-point on interview design — 2026 ML coding interviews increasingly skip rote algorithm recall in favor of realistic production scenarios like today's, testing engineering judgment instead.
Clay closed a $115M Series D led by Wellington Management, an asset manager with roughly $1.4 trillion under management, pushing its valuation from $3.1B last August to $7.1B. This round signals that Wall Street asset managers — not just traditional VCs — now treat AI-agent-driven GTM automation as a predictable, scalable enterprise software category worth betting on.
Euno closed a $23M Series A led by N47, with co-founders of security startups Wiz, Cyera, and Eon joining as angel investors, bringing total funding to $29M. The round signals that the next infrastructure battleground for enterprise AI agents isn't memory capacity — it's a context layer that keeps learning and enforces governance rules in real time.
DeepSeek-V4.1-Flash (`deepseek-flash`) is a 552B MoE with 8B prefill and 16B decode active parameters, a 1M-token context window, and 384K maximum output. It is MIT-licensed, priced at $0.30 input and $1.20 output per million peak-hour tokens, and brings CED, CSA2, and FP4 KV-cache compression to agent workloads.
Nex-N2.5-mini (`nex-agi/Nex-N2.5-mini`) is a 35B MoE with about 3B active parameters, a 262,144-token context window, Apache-2.0 licensing, and a free OpenRouter tier. Nex reports large gains over N2-mini on DeepSWE and Terminal-Bench, with visual-feedback self-correction as its central design goal.
GreyNoise and Blackpoint Cyber independently tracked the same campaign: since August 31, a suspected Russian-speaking actor chained CVE-2026-81578 (auth bypass) and CVE-2026-82078 (unsafe dynamic class loading RCE) in PaperCut NG/MF, built target lists via Netlas.io scanning, then unleashed hundreds of AI agents powered by OpenAI Codex and a DeepSeek model — backed by the Hindsight persistent-memory service and the AionUi multi-agent workspace — to automate Mimikatz, SharpHound, Certipy, Rubeus, and Impacket. The campaign has compromised 440 servers across 395 organizations in 48 countries; one target went from initial access to full domain administrator in seven minutes. Defense: upgrade immediately to PaperCut's Emergency Patch Release 2, and check whether NTDS.DIT has already been exfiltrated via DCSync.
skills is Serverpod's open-source CLI: package authors ship a skills/ directory with their Dart/Flutter package, and users run skills get to auto-install every dependency's Agent Skill into Claude Code, Cursor, Codex, and more. Install: dart pub global activate skills. It addresses the problem of AI assistants not knowing how to use a third-party package without manually pasted docs or hand-written rule files.
This round of LLM & Agent Engineering focuses on the architecture decisions interviewers dig into most in 2026: RAG and agents aren't an either/or choice but a division of labor — RAG handles cited, grounded policy Q&A, while an agent only steps in when the task needs multi-step, cross-system actions. Guardrails split into offline evaluation (catching regressions before release) and online interception (blocking unsafe outputs in real time), governed by one core principle: the model proposes, and a deterministic system authorizes and executes. On evaluation, beyond standard LLM-as-a-Judge scoring, agents also need trajectory-based evaluation that checks each intermediate decision, not just the final answer. Today's practice question is a PracHub-compiled 'design an enterprise support agent' system design prompt, drilling how to separate the RAG path from the agent path and make the authorization layer the centerpiece of your answer.
Mastra @mastra/core@1.65.0 highlights: (1) a new advanced trace query contract implemented across ClickHouse, DuckDB, and Postgres, with bounded time ranges, recursive predicates, and cursor pagination; (2) tenant-scoped batch deletion (up to 1,000 traces per request) that cascades to spans/scores/feedback/metrics/logs; (3) breaking: `@mastra/factory`'s `defineBoard()` becomes a typed phase contract, the global rules object is removed, and two `@mastra/playground-ui` components got renamed slots/props.
Pydantic AI v2.42.0 highlights: (1) a new `GitHubCopilotProvider` lets Agents use GitHub Copilot's OpenAI-compatible API directly as a model backend; (2) `DeferredToolResults.approvals` now rejects invalid values outright — a compatibility change; (3) fixes for Bedrock Converse sampling settings, `$ref` resolution in code-mode function schemas, and lost Anthropic error-recovery state across normalized history.
Cymphony closed a $25M Series A co-led by Sequoia and SMBC Fin Atlas Beyond Fund, bringing total funding — including Sequoia's earlier seed — to $30M at a post-money valuation above $100M. This is Sequoia betting twice on the same team in a market — AI agent identity and data security — that hasn't yet proven itself as a standalone category. The bet isn't on whether the need exists, but on whether it becomes the next must-buy line item in enterprise security budgets.
Harvey closed a $550M round co-led by new investors Diffusion and Lightspeed, pushing its valuation from $11B in March to $15.5B — nearly doubling in nine months. Half the money buys confidence for its own fine-tuned model, half buys agent-security capability via the Guardrails AI acquisition: Harvey is shifting from a law-firm AI assistant into a vertical AI company that trains its own models and manages its own agent risk.
The easiest way to fumble an AI Product Design question is to turn "should we use AI" into a matter of belief, without naming which specific class of task can be automated and which must keep a human in the loop. Today we use a risk-by-confidence matrix to sort tasks into auto-executable, needs-review, and human-required zones, then use progressive delegation to design the cadence at which users actually earn trust in AI, practicing a design prompt about an ops team that won't let AI autonomously run multi-step tasks. The case study is Gusto's AI product Cofounder, whose team publicly explained in September 2026 that AI-moderated interviews are reserved for narrowly scoped, low-risk evaluative research, while depth work always goes to a human researcher.