Laguna S 2.1: MoE 118B with 1M‑token context and “max thinking” tradeoffs for long‑horizon coding
Executive summary Laguna S 2.1 is a 118B-parameter Mixture‑of‑Experts model. Poolside reports it activates roughly 8B parameters per token (~6.8% active), supports up to a 1, 000, 000‑token context, and offers a runtime “max thinking” mode that boosts long‑horizon coding performance at the cost of much higher completion‑token use. Poolside published weights (Hugging Face) and […]
Gemini Flash: Token‑efficient models for agentic automation — benefits, risks, and a 30‑day POC
Google’s Gemini Flash tier: cheaper, token‑lean, and tuned for agentic work, but caveat emptor Think of the Flash tier as swapping a single heavyweight specialist for squads of quick scouts: many smaller, faster model calls instead of one deep, expensive pass. Google/DeepMind published details for three Flash variants on July 21, 2026, Gemini 3.6 Flash, […]
Self‑Distilled Reasoning: Preserving Chain‑of‑Thought in Supervised Fine‑Tuning (Amazon Nova)
Exploring self-distilled reasoning for supervised fine-tuning with Amazon Nova Reported by Amazon (Rushil Anirudh, Shiva Mahalingam, Anupam Dewan), Self‑Distilled Reasoning (SDR) is a practical technique for supervised fine‑tuning (SFT) when your training data lacks chain‑of‑thought (CoT) traces. The idea is simple: have the base model generate reasoning traces, prepend those traces to the supervised targets, […]
Stanford Students Walk Out on Sundar Pichai — A Talent Pipeline Warning for AI Companies
It’s a Modern-Day Draft: Why Stanford Students Walked Out on Sundar Pichai’s Commencement Speech Wired reported that “more than a hundred” students walked out of Sundar Pichai’s July 21, 2026, Stanford commencement. The action was more than a protest against a guest speaker. Organizers framed it as a rejection of the talent pipeline, an elite […]
ChatGPT and Gemini gave inconsistent, often incorrect voting advice in Liberties’ March 2026 test
Liberties’ March 2026 test: ChatGPT and Gemini gave inconsistent, often incorrect voting advice When Liberties (liberties.eu) ran voter profiles that matched Hungary’s winning opposition party Tisza, ChatGPT did not recommend Tisza in 90% of the direct-advice runs. For percentage-matching prompts it assigned Tisza any score in only 2% of cases (Liberties report, March 2026, findings […]
UK datacentre water demand: 3 million litres/day request reveals planning blind spot
One datacentre asked for as much water as 3, 500 homes, and planners didn’t model that demand A proposal to draw up to 3 million litres of potable water a day for a single datacentre (Affinity Water submission to MPs) has pushed a simple question into boardrooms and planning teams: who assumed the country always […]
Meta Ray-Ban Optics Gen-2: Prescription Smart Glasses for Field Work and Enterprise Pilots
Meta’s Gen‑2 Ray‑Ban Optics: prescription smart‑glasses that nudge, not replace, your phone ZDNET wore the second‑generation Meta Ray‑Ban Optics prescription model (the Blayzer Optics) for about six weeks. Their verdict was practical: fit and style are improved, but the hardware and AI updates feel incremental for the price. ZDNET gave the glasses a 4 / […]
Agentic AI for AP: How Tradeshift Replaced Spreadsheets with Amazon Quick Suite and Amazon Q
From slow spreadsheets to an AP copilot: how Tradeshift rewired analytics with Amazon Quick Suite and Amazon Q Month‑end close shouldn’t be a waiting game. Tradeshift found finance teams stuck exporting CSVs, wrestling with six months of retention, and timing out on dashboards that couldn’t handle queries larger than 10, 000 rows. Decisions were made […]
Amazon Quick + NVIDIA NeMo for Supply-Chain Mitigation: Ranked, Evidence-Backed AI Agent Workflow
TL;DR Turn supply‑chain alerts into ranked, evidence‑backed mitigation plans by pairing an Amazon Quick workspace (business UI and chat) with an NVIDIA NeMo Agent Toolkit workflow running behind AgentCore. The pattern separates a governed, lightweight front end from a testable, observable decision engine. The sample repo includes CloudFormation, scripts, sample data, and an eval harness. […]
Stream Deck+ Recreates Codex Micro: Map ChatGPT Agent Controls, Costs and Trade-offs
TL;DR David Gewirtz rebuilt most of OpenAI’s Codex Micro control surface using an Elgato Stream Deck+ and a handful of plugins, icons, and hotkeys. The result is a cheaper, more flexible multi‑app setup that reproduces agent switching, accept/decline, branching, mic toggle, dial behavior, and pulsing status lights, with trade‑offs around native status integration, tactile keycaps, […]