AMD shares jumped 9.6% to a record $613.31 on Monday, pushing its market cap above $1 trillion for the first time and making it the fourth US chipmaker to cross the threshold, after Nvidia, Broadcom, and Micron. The stock has gained roughly 185% in 2026, far outpacing the Nasdaq's 15.8%, as AMD increasingly positions itself as a full AI computing infrastructure competitor to Nvidia, selling complete systems combining processors, networking equipment, and related hardware rather than standalone accelerator chips.
Factories installed 603,000 industrial robots worldwide in 2025. China accounted for 354,000. The global installed base passed five million, showing how concentrated the latest wave of factory automation has become.
LTT Labs measured the A20 Pro at 55% above the iPhone 17 Pro's A19 Pro in one 3DMark graphics test. Its broader review also tests sustained performance, display brightness, charging, and battery life.
Legged robots burn energy differently than wheeled ones, constantly working at the joints just to hold body weight up and losing kinetic energy every time a foot strikes the ground, so KAIST's team built RAIBO2 by comprehensively modeling every source of energy loss instead of optimizing pieces in isolation. The gains came from stacking hardware and software fixes: an optimized mechanical structure, a redesigned motor driver circuit that cuts electrical losses in the power-control stage, and a locomotion policy trained via RL specifically to minimize energy dissipation.
Fault-tolerant quantum computers need a classical decoder to keep pace with error-correction cycles in real time, or a backlog builds that slows the whole computation exponentially. IonQ researchers showed a decoder running on just twelve cores of an Apple M4 Max MacBook Pro chip could keep pace with a simulated 408-logical-qubit trapped-ion machine, using two sliding-window decoders per code block, a slower, highly accurate one tracking corrections for the whole computation, and a faster one handling time-critical logical measurements.
Dexterous manipulation needs both long-term foresight and fast reactive control, but existing models tend to be good at only one, so LiMA decouples the two into an asynchronous dual-system hierarchy: a slow system handles sparse long-horizon intent, a fast system handles dense high-frequency motion, bridged by a Latent Schrödinger Bridge Coupling mechanism that treats aligning the two as a probabilistic transport process rather than a direct mapping. This cuts inference latency by 45.8% versus Cosmos-Policy, and across six bimanual manipulation tasks, LiMA hit a 70.8% overall success rate while holding up on scenarios it wasn't trained on.
Semiconductor design spans dozens of specialized tools - RTL, synthesis, place-and-route, timing sign-off, and DFT - each written by a different vendor with different assumptions about what "done" means. Every handoff between tools risks silently losing design intent, a problem serious enough that first-time silicon success has fallen sharply, with most respins still traced to these intent-transfer failures rather than raw errors. The consensus emerging is "evidence-driven automation": AI can propose fixes, extract properties, and orchestrate handoffs between agents, but every output has to remain independently checkable through formal proof, deterministic verification, and audit trails, since an AI-generated assertion is a hypothesis, not proof, until it's checked against a real engine.
Meta introduced VR Glasses, a headset weighing about 100 grams, five times lighter than Meta Quest 3, achieved by splitting the system in two: the glasses handle sensors and a 5K micro-OLED display at 37 pixels per degree through custom pancake lenses, while a separate puck connected by an optical tether handles compute, battery, and storage, clipping to a pocket or bag. The puck runs Qualcomm's new Snapdragon Reality Elite processor and delivers up to three hours of continuous high-resolution playback on 45W fast charging, while the glasses integrate Dolby Vision, Dolby Atmos spatial audio, and hands-free control through Meta AI, eye tracking, and natural gestures with no controllers.
Alibaba's T-Head semiconductor unit unveiled the Zhenwu V900, which CEO Eddie Wu called the most powerful AI chip in China, delivering roughly 3x the performance of its predecessor, the M890. A single cluster built on the V900 can reportedly link up to 500,000 cards for frontier model training and inference, the kind of scale needed to support the 5-10 trillion-parameter model Alibaba's Qwen team is now targeting.
Microsoft opened a 15,000-square-foot quantum research center at the University of Maryland's Discovery District, giving DARPA on-site, hands-on access to a topological quantum system built on its Majorana 2 chip, a real shift from the remote evaluations DARPA previously ran from Microsoft's Redmond or European facilities.
China's Kinetica 1 Y18 rocket launched nine satellites on September 20, headlined by Supercomputing-1 (S-AIDC-1), which pairs a high-resolution optical payload with an onboard AI computer to process Earth-observation imagery directly in orbit rather than beaming raw data back to ground stations first, cutting cross-regional processing time from hours to minutes.
A new market report projects the US silicon photonics market growing tenfold from roughly $1 billion in 2025 to $10.76 billion by 2035, driven by AI data centers pushing past what traditional electrical interconnects can handle.
Lumentum, Qualcomm, and Corning are jointly demonstrating a board-level optical die-to-die interconnect at ECOC 2026, aimed at extending high-bandwidth, low-latency chip-to-chip connectivity well beyond what copper traces can support in AI systems.
Supermicro has started shipping Nvidia's next-gen Vera Rubin NVL72 systems bundled with its Data Center Building Block Solutions, integrating compute, direct liquid cooling, power delivery, storage, and networking into one deployable package rather than selling GPUs alone.
Sponsored
Everyone can ship fast. The challenge is shipping reliably, securely, and autonomously, with minimal human involvement or none at all.
Join Dr. Ariel Shiftan, Co-Founder & CTO of Autonomous Security [5], for an inside look at the architecture behind their autonomous software factory:
* How AI agents move from tasks to end-to-end delivery * The risks of long-running AI workflows, and how to keep them
* The KPIs and trade-offs that matter when AI, not humans, drives development
See how they built it, what they learned, and how it changed the way they develop software.
Graphalgo malware appeared in two Terraform providers and two Go modules. Trigger conditions hide execution until targeted inputs appear, and the payload reports host data through Slack, receives encrypted commands through Slack and an Arbitrum Sepolia contract, then runs Go or Node code. Aikido recommends removing affected packages, isolating exposed hosts, rotating credentials, reviewing deployments, and reimaging machines.
GitLab incoming email addresses embed non-expiring tokens that act as the owner across accessible projects. A leaked address can create merge requests, submit patches, or push code to private projects after changing project identifiers, and can be used to bypass IP restrictions via email. Aikido recommended that organizations reduce their attack surfaces by preemptively rotating the access tokens in their GitLab email addresses and by scanning for these addresses in their development environments and repositories as if they were any other secrets.
Someone registered third-party[.]com, a non-IANA-reserved placeholder cited in 1,700+ GitHub repos, including AI agent skills and MCP docs, and it now serves a ClickFix PowerShell lure only to Windows visitors, while yoursite[.]com and your-domain[.]com push scams to macOS users that static scans miss.
A bug bounty hunter built PhantomRaven, an LLM-generated JS info-stealer, and pushed it through typosquatted npm packages that used HTTP-URL dependencies to fetch payloads on install. It grabs OS/hostname/IP data, Git/npm credentials, and CI/CD env vars from GitHub Actions, GitLab CI, Jenkins, and CircleCI, then sends data via HTTP GET/POST. The operator used compromises to file bug bounty claims on HackerOne, Bugcrowd, Intigriti, YesWeHack, and HackenProof rather than selling logs. Npm 12+ now blocks preinstall scripts by default. Use --ignore-scripts and a private registry to block this vector.
At DEFCON 34, Intigriti researchers reported on various ways to abuse AI customer service agents. This article walks through different methods to abuse the agents to send phishing emails, invoke tool calls, and bypass MFA. It also demonstrates mechanisms to steal third-party OTPs and poison knowledge base articles to deceive victims.
CARBONATO worms through unauthenticated Docker daemons on port 2375 via privileged, host-mounted containers, then installs the stock open-source Hermes Agent and overwrites its SOUL.md persona so that Telegram-tasked LLM sessions run post-exploitation and loot AI API keys before SSH credentials. ThreatDown recovered the toolkit from the crew's own open registry, which doubles as the fleet's update server, and ties the operators to Costa Rica through voseo Spanish, UTC-6 build timestamps, and a +506 Telegram handle. Keep the Docker API off the network, hunt for a SOUL.md containing "GH0ST" rather than blocklisting the legitimate hermes-agent package, and inventory and rotate AI API keys like bank credentials.
Sponsored
ThreatSpike tried TLDR after Google and LinkedIn ads fell short. Its campaigns now generate $25K–$75K in new business pipeline each. If your company markets a cybersecurity product, share this case study with your marketing team [14].
Island runs an enterprise browser and SASE platform combining ZTNA, device posture checks, and inline DLP to block unapproved clipboard, download, and screen-capture actions.
ceasta is a single C++ tool that combines an IDA-style disassembler, a pseudocode decompiler, a function graph, and an x64dbg-style debugger for PE, ELF, and raw shellcode, with a full GUI on Windows and a CLI with a ptrace debugger on Linux. Its built-in MCP server (ceasta-cli mcp target.exe) lets Claude Code or Cursor decompile, rename, and diff builds. With --allow-debug, it can also set breakpoints and call functions in the running process. The decompiler is best-effort, with no types or structs yet.
Pixi Audit, now in beta, scans Pixi lock files for vulnerable conda-forge and PyPI packages, and --fix bumps direct dependencies to the lowest compatible fixed version before re-running pixi update. Matching relies on Basilisk, prefix.dev's new conda-forge vulnerability database, which links repackaged conda-forge builds to their upstream PURL and CPE identities so existing OSV and NVD advisories apply. For CI, --severity high sets the failure threshold, with JSON and SARIF output, though it doesn't cover statically linked binaries yet.
An OpenAI agent gained unauthorized access to a Medicare statistics portal in June. It accessed public and non-public files, including aggregate health data and internal filenames. Australia was notified on September 10 through a generic disclosure mailbox. Officials later said agents also targeted two state-government sites.
Elsevier.com, Evolve.elsevier.com, and Submit.elsevier.com redirected visitors to a LAPSUS$ “Chapter II” page for at least 78 minutes on September 21. Researchers suspect a DNS, CDN redirect rule, or management account change. Elsevier restored service and said its investigation found no compromise of core platforms, customer data, research content, or operational systems.
Mandiant published a five-pillar blueprint (endpoint, repositories, artifacts, CI/CD, and deployment) for hardening the SDLC, prompted by campaigns that poison GitHub Actions caches, extract OIDC tokens, and repoint mutable action tags to publish malicious packages that still carry valid provenance. Standout recommendations include a seven-day cooldown before new public package versions become installable, pinning images and actions to digests and commit hashes, ephemeral runners with egress limited to approved registries, and replacing PATs with GitHub Apps and hardware-backed SSH keys. It also addresses AI coding agents, advising teams to keep .env files out of model context and require human review of AI-generated code.
Attackers spoofed Astrana Health's main phone number and impersonated staff to access company servers.
Canonical is replacing Ubuntu's 4-week regular and 2-week security kernel SRU cycles with overlapping 2-week cycles that ship a kernel every week.
CISA has flagged CVE-2026-63077, an unauthenticated RCE in JetBrains TeamCity On-Premises reachable through the agent polling protocol, as exploited in ransomware attacks.
Sponsored
In the 2026 Gartner Critical Capabilities for SOAP, Redwood scored highest in 4 out of 5 use cases…and tied for first in the last one.
Now you can see why. Download the Gartner report [5] for free to see the full score breakdown.
👉 With 30 years of automation expertise, RunMyJobs [6] empowers IT teams to speed up automation design, deploy mission-critical business processes and reduce total cost of ownership (TCO).
Over 4 billion SaaS executions run annually across Fortune 50 and global enterprises.
2026.09.25·redwood.com·Sponsor·TLDR IT Anthropic committed $11.6 billion over seven years to Akamai's cloud infrastructure to support growing CPU workloads. The agreement could expand by another $9 billion and includes warrants that could give Anthropic roughly a 5% stake in Akamai.
2026.09.25·akamai.com·5 min read·TLDR IT Amazon Web Services has become the first hyperscaler to receive NATO Restricted approval, allowing member nations and defense partners to handle mission-critical restricted workloads across 15 qualifying AWS Regions. Evaluated by Spain's National Cryptographic Centre, the accreditation provides a pre-assisted security baseline and streamlined access to cloud capabilities for the alliance.
2026.09.25·itpro.com·2 min read·TLDR IT Databricks acquired Row Zero to bring a native spreadsheet interface to Genie, letting business teams analyze live enterprise data with built-in permissions and auditability. The planned integration supports familiar formulas and pivot tables while reducing reliance on exported files that lose governance controls.
2026.09.25·databricks.com·4 min read·TLDR IT Foundation models face the same tribal knowledge challenges as human workers rather than requiring specialized database architectures. Advances in general foundation models will continue to subsume custom agent harnesses, making standard software documentation practices the primary solution for semantic context.
2026.09.25·emptysqua.re·6 min read·TLDR IT UiPath announced Cartographer and the Map of Work at its FUSION 2026 event, introducing a new product designed to capture enterprise processes and generate governed process maps for AI agents and automations. The tool uses AI to absorb existing documentation, interview experts, and track decisions through a ledger to keep process maps continuously updated.
2026.09.25·diginomica.com·7 min read·TLDR IT Cognition has launched native support for @ mentioning Devin in Microsoft Teams alongside a first-party Microsoft 365 integration. The update allows Devin to participate directly in Teams chats, channels, and threads, as well as connect to mail, calendar, files, tasks, and directory services through Microsoft 365 MCP servers. The integrations feature delegated per-user permissions and resource-specific consent controls designed for security review.
2026.09.25·devin.ai·4 min read·TLDR IT Docker introduced managed cloud sandboxes that let coding agents keep working after developers disconnect, with a command to move sandbox files between local and cloud environments. Each agent runs in its own microVM, with configurable network access and usage-based compute pricing starting at $0.07 per hour.
2026.09.25·docker.com·6 min read·TLDR IT Canonical is replacing Ubuntu's existing kernel update cadence with overlapping two-week cycles that produce a release every week. The change aims to shorten patch delays as AI-assisted bug discovery and broader Linux CVE reporting increase the vulnerability backlog.
2026.09.25·theregister.com·3 min read·TLDR IT Pinecone's Bring Your Own Cloud offering is generally available across AWS, Google Cloud, and Azure, keeping customer data and queries inside customer-controlled accounts. Pinecone manages upgrades and scaling through an outbound, pull-based mechanism without requiring standing inbound access to the environment.
2026.09.25·pinecone.io·4 min read·TLDR IT Kubernetes has fixed a medium-severity vulnerability designated CVE-2026-76654 in the Windows kubelet that could force NTLM authentication to an attacker-controlled network share.
2026.09.25·freenode.net·1 min read·TLDR IT Ando is building team messaging designed for employees and AI agents to collaborate.
2026.09.25·techcrunch.com·4 min read·TLDR IT Sponsored
94% of B2B buyers now use AI somewhere in their buying journey, so prompt tracking was a reasonable first move. Build a prompt list, get a score. Ask the same question twice and the answer comes back different. It tells you that you dropped. Not why, not which page, not what to fix.
Join GrowthX CEO Marcel Santilli on Oct 22: WHAT 10,000 HOURS OF AI VISIBILITY TAUGHT US [5]. Five lessons from 2.6 million AI responses, plus the playbook.
Free, virtual, replay for everyone who registers!
23% of US consumers, or roughly 61 million people, now start retail research with AI. Average AI-assisted spending rose 17% from $889 in June to $1,039 in August. Yet 60% of AI-assisted purchases still close on Amazon. In a controlled test, 40% of consumers chose a marketplace even when told other retailers offered the same product, price, shipping, and returns.
McDonald's is piloting a media network across 450 US company-owned restaurants, selling ad space on its app, kiosks, and menu boards, with the goal of building a billion-dollar business. Its marketing is shifting to fewer, bigger global campaigns built on its own IP. It is roughly doubling spending on long-running programs about its brand and food quality while cutting short-term promotions.