Darren's Daily Briefing

Cloud-to-orbit CTO strategy, satellite SaaS and platforms, end-to-end space security, agentic AI & MCP, space technologies, and deep technical research.

CTO Topics — 1 article

Kora Splits Local AI Across CPU, GPU, NPU and Nearby Machines

Tom's Hardware · November 11, 2026
Market
Hybrid local-to-cloud inference and heterogeneous edge compute / platform architects placing AI across constrained user, ground and eventual flight-class systems.
Trend
This is a vendor prototype and 2027 product plan, not a production deployment. Kora reports Gemma 4 26B at 25 decode tokens/s and 550 prefill tokens/s with a 32K context on Snapdragon X2 Elite's Hexagon NPU, and projects 70% of managed model calls will run locally; the exact chip, RAM configuration, pricing and independent benchmark are missing.
Tech Highlight
The Linux OS uses a hybrid inference layer to place models across CPU, GPU, NPU and LAN peers, then falls back to cloud models. Its current Qualcomm path uses GGUF plus llama.cpp's Hexagon backend and mainline FastRPC, while planned AMD and Intel paths use different runtimes and reserve their NPUs mainly for embeddings or speech—evidence that portable abstractions still terminate in vendor-specific hardware/software partitions.
6-Month Outlook
Watch for reproducible power, memory and latency measurements on shipping X2 Elite PCs, a documented cloud-fallback policy, and working AMD/Intel backends. Comparable results across vendors would validate a transferable edge control plane; a Qualcomm-only launch or opaque placement decisions would expose lock-in behind the hybrid claim.

SaaS and Platform Tech Markets — 1 article

NATO Tests a Multi-Orbit SATCOM Platform Model Without Presuming Service Equals Assurance

NATO Communications and Information Agency · November 13, 2026
Market
Integrated military-commercial SATCOM as a managed platform / NATO headquarters, deployed forces, ships, aircraft and manpack users.
Trend
This is a draft request for information, not an award, prototype or deployment. NATO is testing a two-speed operating model: adopt commercial services where they meet requirements, retain military capacity where assurance is indispensable, and integrate both through open interfaces while explicitly rejecting provider count or an ordinary commercial contract as proof of assured availability.
Tech Highlight
The proposed control plane spans GEO, MEO, LEO and HEO capacity; multi-band terminals; DIFI-based virtualized ground interfaces; open APIs; NATO-controlled keys; and automated monitoring, path selection and capacity allocation with human oversight. Respondents must separate deployed capabilities from prototypes, disclose third-party dependencies, and itemize capacity, terminal, integration, licensing and continuity costs.
6-Month Outlook
Watch the RFI become a funded procurement with numerical SLOs, accreditation boundaries, failover exercises and portable interfaces across at least two providers and orbits. A test showing restoration under interference or gateway loss would validate the platform thesis; a prime-led bundle without measurable portability would preserve lock-in.

Security + SaaS + DevSecOps + AI — no new items today

No eligible articles in the last calendar month. Check back tomorrow.

Agentic AI & MCP Trends — 1 article

AgentCore V2 Restores Isolated Agent Sessions from Snapshots

Amazon Web Services · November 13, 2026
Market
Durable and interactive agent runtimes / teams operating code, claims and voice workflows in isolated serverless sessions.
Trend
This is a production architecture guide, not a comparative benchmark. AgentCore V2 targets roughly two-second session starts by building an environment once and restoring a snapshot for each isolated microVM; AWS says unused memory is reclaimed for usage-based billing, but publishes no fleet-wide latency distribution or cost delta here.
Tech Highlight
At deployment, the runtime starts the container, waits up to 120 seconds for GET /ping, clears the OS file cache, freezes memory and disk, and stores the snapshot. Each session resumes from that point with imports, objects and clients initialized, while large files load lazily; build hooks can pre-stage browsers or CLIs, and sub-second experiences still require pre-warming on user engagement.
6-Month Outlook
Watch customer P50/P95 cold-start and cost-per-completed-workflow results, plus evidence that snapshot rebuilds and lazy file faults remain predictable for long-running agents. Consistent starts under two seconds with lower idle memory would validate the model; stale state, slow snapshot creation or heavy first-read penalties would shift complexity rather than remove it.

Space Technologies — 1 article

Vulcan Failure Destroys 40 Amazon Leo Satellites

Spaceflight Now · November 15, 2026
Market
Launch availability and constellation deployment resilience / Amazon Leo, ULA, national-security customers and satellite-network schedulers.
Trend
This is a flight failure and the first outright ULA mission loss since the Boeing-Lockheed joint venture formed in 2006. Vulcan suffered an anomaly about ten minutes after liftoff and failed to orbit all 40 production satellites on its fifth mission and first flight in nine months, converting launcher concentration directly into constellation schedule and capacity loss.
Tech Highlight
The six solids appeared nominal, while the methane BE-4 first stage appeared to burn longer than the preflight timeline before the mission failed after staging; ULA has not identified a root cause. Transferable architecture value therefore lies in telemetry-driven fault isolation and manifest diversity, not premature attribution: Amazon can absorb identical satellite loss more readily than it can instantly replace a grounded launch lane.
6-Month Outlook
Watch ULA publish a causal chain and corrective action, regulators clear return to flight, and Amazon re-manifest satellites across Vulcan, Ariane 6 or other providers. A bounded cause with a successful repeat mission would contain the failure; prolonged grounding or launch-provider reshuffling would reveal the true unit-economic and schedule cost of deployment concentration.

Deep Technical & Research — 1 article

llama.cpp 1.0 Makes API Stability a Local-Inference Contract

ggml-org / Hugging Face · November 13, 2026
Market
Portable local and edge inference runtimes / device, ground-edge and disconnected-system developers.
Trend
This is an open-source platform release and compatibility policy, not a new performance benchmark. After 3.5 years of development, llama.cpp and ggml reached 1.0 and now promise major-version bumps for public C API or ABI breaks, addressing lifecycle risk for downstream products that embed quantized multimodal inference.
Tech Highlight
libllama and libggml expose versioned public headers while keeping portable C/C++, quantization and optimized backends across Linux, macOS, Windows, Android and iOS. CI spans self-hosted hardware from NVIDIA, AMD, Qualcomm, Arm, IBM and Intel; this improves software portability, but it does not establish deterministic performance, radiation tolerance or flight qualification on any specific SoC.
6-Month Outlook
Watch downstream projects adopt pinned 1.x interfaces, distributions ship reproducible packages, and compatibility tests catch backend regressions across Arm and accelerator targets. Fewer forced rebuilds with stable accuracy and throughput would make llama.cpp a credible edge substrate; frequent major bumps or divergent vendor backends would weaken the contract.