THE DOUG LEONE PLAYBOOK
Doug Leone, after 26 successful years at Sequoia, says he will consider retiring if he feels irrelevant in nine months.
2026.09.04
Search the archive
Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Results
27,290 matchesPRODUCT OR DISTRIBUTION: WHICH ONE ACTUALLY BUILDS A COMPANY?
People love to say build something people want, but a seed fund estimates that about 5% of its early-stage portfolio has a demand problem, while the other 95% have a customer-acquisition problem.
2026.09.04
UBER AND WAYVE LAUNCH LONDON'S FIRST PUBLIC ROBOTAXI SERVICE, BEATING WAYMO TO THE CITY
Uber began deploying Wayve-powered autonomous vehicles across London on Thursday, marking Wayve's first commercial public service anywhere in the world and making London only the second European city with robotaxis, after Zagreb. The initial fleet, fewer than 20 Ford Mustang Mach-Es with plans to add Nissan Leafs later, keeps a licensed safety operator behind the wheel for now, and riders booking UberX, Comfort, or Electric may get matched with one at no extra cost.
2026.09.04
FIGURE SECURES UP TO 100,000 GPUS FROM NSCALE TO TRAIN ITS HUMANOID ROBOTS
Figure signed a deal with Nscale to deploy up to 100,000 Nvidia Vera Rubin GPUs, starting in Barstow, Texas, in the second half of 2027, an initial $3.5 billion compute commitment with intent to scale past $6 billion, with Nscale also making a strategic investment in Figure. It follows last week's launch of Index, Figure's crowdsourced humanoid training dataset now generating 35 minutes of data every second, addressing the other half of the bottleneck: Figure says it's now largely constrained by data and compute rather than robot hardware itself in training its Helix model.
2026.09.04
MICRON REPORTEDLY EXPLORES NEAR-GPU NAND, A FLASH TIER SITTING BETWEEN HBM AND THE SSD
Micron is reportedly investigating near-GPU NAND, high-endurance flash modules placed right on the GPU package where HBM sits today, aiming to fill the massive latency and bandwidth gap between HBM and a system's SSD. The pitch is memory-bound inference: large models buy extra accelerators purely to hold weights and KV cache, an expensive way to add capacity since HBM is the priciest, tightest-supplied memory around, so a cheaper flash tier fast enough to keep compute fed could let the same model run on fewer GPUs.
2026.09.04
WIRED'S FIRST LOOK AT NVIDIA'S RTX SPARK LAPTOPS, ARRIVING FALL 2026
Nvidia's RTX Spark superchip fuses a 20-core Grace Arm CPU with a Blackwell RTX GPU on one 3nm package, delivering up to 1 petaflop of AI performance and 128GB of unified LPDDR5X memory, enough to run 120-billion-parameter models locally with up to a 1-million-token context window. It's Nvidia's first true client SoC for Windows, pairing native CUDA support with full RTX graphics tech, so the same chip targets AI development, creative work, and gaming. Jensen Huang promises at least 100fps at 1440p in modern AAA titles.
2026.09.04
BIMO, AN OPEN-SOURCE BIPEDAL ROBOTICS KIT WITH SIM-TO-REAL TRANSFER
Bimo is an open-source, 3D-printable bipedal robot kit built with custom quasi-direct drive actuators, ROS 2 integration, and simulation-ready environments for reinforcement learning locomotion policies. The repository provides complete mechanical models, electronics schematics, and training pipelines to enable reliable sim-to-real transfer on a desktop-sized robot. It offers an accessible, affordable testbed for experimenting with legged locomotion dynamics, custom motor control schemes, and modern reinforcement learning algorithms without requiring high-cost industrial hardware.
2026.09.04
ARDUINO WITH ZEPHYR
Arduino trained a generation of young engineers, and now, post-Qualcomm acquisition, they are professionalizing the framework. The recent Arduino Core on Zephyr 1.0.0 release bridges the familiar, beginner-friendly Arduino API with the industrial-grade Zephyr Real-Time Operating System. This integration allows developers to leverage enterprise capabilities like preemptive multi-threading, advanced power management, and standardized networking stacks while writing standard Arduino code.
2026.09.04
NVIDIA'S PAIR TURNS SPARE RTX GPUS ON YOUR HOME NETWORK INTO ONE INFERENCE CLUSTER
As multi-agent AI workflows spawn dozens of independent model calls from a single task, all competing for one local GPU's execution queue, Nvidia's new open-source PAIR router spreads those calls across every eligible RTX GPU, RTX PRO workstation, DGX Spark, or Apple M4+ device already sitting on a home network. It doesn't split one inference request across multiple GPUs or pool VRAM into a bigger virtual accelerator - each request still runs whole on one node, but it proxies existing Ollama and LM Studio interfaces so agents keep using the connection they already expect while PAIR handles discovery (via mDNS), MTLS-secured pairing, and live scheduling based on which paired node is actually ready.
2026.09.04
DYSON'S CAMERAJET TOOTHBRUSH USES A 100K PIXEL CAMERA TO AUTO-TARGET AND JET-FLOSS GAPS BETWEEN TEETH
Dyson's CameraJet packs a 100,000-pixel macro lens camera into the brush head, running its "Gap Optical Targeting" machine learning system to detect and predict gaps between teeth in real time, triggering a jet of water within 100 milliseconds of spotting one.
2026.09.04
TSMC'S EQUIPMENT PROCUREMENT FORECAST NEARLY DOUBLED IN SIX MONTHS AS IT BUILDS 20 FABS AT ONCE
TSMC Senior VP Jeffrey Hou revealed at SEMICON Taiwan 2026 that AI-driven demand has pushed the company's equipment purchase forecasts to nearly double what they were just six months earlier, as it simultaneously builds roughly 13 fabs in Taiwan and 5-6 overseas, 4-5 times its historical pace.
2026.09.04
AGS-PLANTSEG - A DATASET AND BENCHMARK FOR AGRICULTURAL ROBOTICS
AGS-PlantSeg is an open-access 3D point cloud dataset and benchmark designed for semantic segmentation in complex agricultural environments across varying plant growth scales.
2026.09.04
HTC'S $499 VIVE EAGLE SMART GLASSES LAND IN THE US, PITCHING A PRIVACY-FIRST DESIGN AGAINST META
HTC's Vive Eagle glasses go on sale in the US, Europe, and Australia at $499 (or $599 with photochromic lenses).
2026.09.04
BROADCOM'S Q4 FORECAST MISSES ESTIMATES, FALLING SHORT OF NVIDIA'S AI BENCHMARK
Broadcom guided to $34.8 billion in fourth-quarter revenue, below the $35.1 billion analysts expected and well under some estimates that topped $36 billion, sending shares down about 4% after hours.
2026.09.04
ATTACKERS EXPOSE ONGOING AI TOOL USE TARGETING ORGANIZATIONS IN LATIN AMERICA
Two intrusion clusters targeting Latin American critical infrastructure accidentally exposed their operations by leaving a self-hosted NextChat instance open to the public on TCP 3000. The exposed prompt history and staging scripts revealed a heavy reliance on AI tools to rapidly iterate through failed SAM and NTDS.dit credential dumps using volume shadow copy workarounds. To detect this infrastructure, defenders should actively hunt for the m-doxa on duckdns[.]org subdomains and monitor for multi-SAN Let's Encrypt certificates rotated between February and June.
2026.09.04
MALICIOUS .GIT CONFIGS CAN MAKE CLAUDE, CODEX, CURSOR, AND OTHER AI AGENTS RUN ATTACKER CODE
Manifold Security found eight flaws in seven coding agents. A received repository can use .git/config to trigger commands through core.fsmonitor, and the commands run with the developer's account permissions, outside sandbox controls and without approval. Fixes exist for goose, Codex, and one Claude Code path. Hermes Agent, Qwen Code, Grok Build, and another Claude Code path remained exposed.
2026.09.04
CODER'S REGISTRY INFRASTRUCTURE COMPROMISED TO PUSH MALICIOUS MODULES
An attacker added unauthorized IPs to Coder's Cloudflare pool for registry.coder.com and served credential-stealing Terraform modules between 07:35 and 21:45 UTC on August 31, exfiltrating cloud, CI/CD, OIDC, and SSH secrets to coder-infra[.]com. Patched releases are 2.37.0, 2.36.4, 2.35.7, and 2.34.9.
2026.09.04
ROGUE SCREENCONNECT INSTALLATIONS ACROSS UNRELATED HOSTS SUGGEST WORM-LIKE ACTIVITY
Modified ScreenConnect clients recovered by Huntress watch the EndPointStatusMessage.Connections collection for newly observed Host sessions, then register 1.vbs through 4.vbs with ScreenConnect's virtual file-transfer system and queue them with the action set to Run, so merely connecting to an infected client can hand the Host the same four-stage VBScript chain. The loader profiles first: 1.vbs writes a three-bit state to %TEMP%\value.txt covering existing ScreenConnect installs, whether Defender is the only endpoint protection present, and whether RAM exceeds 5GB, and that value selects a user-level backdoor, an access package with a ComputerDefaults.exe UAC bypass and AMSI tampering, or a combo branch carrying wstunnel disguised as Themes.exe and XMRig renamed SearchIndex.exe. Defenders should treat RunFiles or RanFiles entries in ScreenConnect server audit logs showing those scripts executed from Process: Guest as indicators of compromise, check for the WindowsServiceHost Run key pointing at WindowsServiceHost.vbs in AppData, and note that Huntress recommends reimaging affected hosts rather than cleaning them.
2026.09.04
FROM PATCH TO EXPLOIT; USING CLAUDE CODE TO REVERSE ENGINEER A ZERO-DAY IN PAPERCUT NG
Kev Breen used Claude Code against a patched PaperCut NG zero-day, spotting SQL injection IoCs in error logs and finding the patch blocked CALL/EXEC/SYSCS_ calls in the SQL lookup template. Claude decompiled the JARs, found an auth-bypass filter flaw using a matrix parameter, then a second bypass in pageValidate letting an unauth service parameter reach admin pages. Chaining both gave unauthenticated RCE via arbitrary file write as the papercut user, confirmed with a working Python PoC. A guardrail swapped the model mid-session when it probed exploitation logs. Breen later found a third bypass defeating the emergency patch and reported it to security@papercut.com, who said they were already aware.
2026.09.04
THE OUTSIDER PHISHING KIT: A RESILIENT THREAT IN THE FACE OF LAW ENFORCEMENT ACTION
Despite recent takedown efforts by Google and the FBI, the ChenLun Outsider Phishing-as-a-Service platform has proven highly resilient, with over 700 new phishing pages deployed in a single month. The platform operates a sophisticated kill chain that utilizes WebSockets to stream victim keystrokes in real time while allowing affiliates to manually push contextual 2FA prompts on demand. Defenders can hunt for this infrastructure across all known templates by looking for the unique GetSyncSetting configuration file and an alphabetical page prefix convention that maps directly to the victim exploitation stage.
2026.09.04
AIR SECURITY (PRODUCT LAUNCH)
AIR Security launched a firewall that inventories and evaluates skills, plugins, MCP servers, and add-ons. It detects malicious instructions, hidden behavior, excessive permissions, typo-squatted packages, and risky updates, then revokes access across dependent workflows.
2026.09.04
air.securityTLDR InfoSec
LEASH
The open-source safety and control layer for personal AI agents: Engine, Desktop, Mobile, proxy, and shared contracts. Leash sits between your AI agents and the actions they take. Safe work keeps moving. Sensitive work becomes a clear decision you can understand and control.
2026.09.04
AWSHOUND: AN OPENSOURCE AWS OPENGRAPH COLLECTOR
AWSHound is a self-hosted open-source collector that maps AWS environments into BloodHound OpenGraph datasets to visualize attack paths. Unlike existing mapping tools, AWSHound performs strict offline IAM evaluations by intersecting identity policies with permissions boundaries and SCPs to ensure relationships are only drawn where effective permissions resolve to an explicit allow. Operating securely through standard AWS SDK credential chains, the tool demonstrated massive scalability during internal testing by mapping a 500-account organization into 10.5 million edges in approximately 15 minutes.
2026.09.04
UK MOVES TO BLOCK HIGH-RISK TECH SUPPLIERS FROM CRITICAL INFRASTRUCTURE
UK ministers tabled amendments to the Cyber Security and Resilience Bill on August 24. The changes let them block high-risk technology suppliers from critical sectors. The move followed an Iran-linked attack that shut a UK energy facility for four days. The bill already mandates tight incident reporting and penalties, and SMEs supplying critical operators may face exclusion without adequate controls.
2026.09.04
OPENAI'S NEW REASONING TECHNIQUE ALARMS AI SAFETY EXPERTS
OpenAI's Astra model uses "recurrent depth," letting it loop through the same query multiple times instead of reasoning step by step, which makes chain-of-thought logs harder to read. Scaling the technique could push reasoning fully into latent space, killing visibility into model behavior. OpenAI's Jakub Pachocki said Astra's chain of thought stays legible and that monitoring remains a research priority.
2026.09.04
AI AGENTS HELP COMPRESS RANSOMWARE INTRUSION TO UNDER 10 HOURS, RAISING STAKES FOR CISOS
A ransomware operator crossed an enterprise network in under 10 hours using AI agents, work Unit 42 estimated would have taken human hands roughly two weeks. The 50-plus ATT&CK techniques observed were unremarkable on their own, and analysts quoted here read the intrusion as human-directed, with AI handling delegated tactical work rather than anything approaching autonomy. The more durable observation is about transitive authority, where a repository account with no cloud admin rights can still alter a workflow that assumes a privileged cloud role, which is the gap that made speed matter.
2026.09.04
VALLEYRAT BACKDOOR DISGUISES ITSELF AS ADWARE
Kaspersky found ValleyRAT hiding inside a modified build of the Chinese adware QN Wallpaper, sideloading a malicious libcef.dll through the signed QnWallpaper.exe to gain a trusted parent process that users routinely whitelist.
2026.09.04
RUSTBOT, THE MACOS MALWARE USED IN THE LATEST RUST SUPPLY CHAIN ATTACK
Poisoned crates.io releases of arrayref, internment, and append-only-vec pulled in a malicious build-time dependency that fetched an arm64 Mach-O implant during cargo build, harvesting ~/.aws, ~/.ssh, and ~/.kube credentials before its RSA tasking channel was neutered by the private key shipping inside the binary.
2026.09.04
MICROSOFT TEAMS, OUTLOOK FAIL TO LAUNCH ON ARM-BASED WINDOWS PCS
Teams and New Outlook can crash or fail to launch on ARM-based Windows 11 devices after recent security updates, particularly on newly imaged machines missing Microsoft Store updates. Microsoft recommends updating the Auto Super Resolution Package as a temporary workaround while it prepares a permanent fix.
2026.09.04
SALESFORCE SIMPLIFIES ITS EDITIONS BY BUNDLING SLACK, SECURITY, AI, AND ANALYTICS
Salesforce has introduced new Core, Advanced, and Max editions that package Agentforce, Slack, Tableau Next, security capabilities, and support into fewer SKUs. The move reflects a broader push by SaaS vendors to consolidate products into larger platform bundles rather than selling each capability separately.
2026.09.04
DEFYING GRAVITY: THE SYSTEM OF RECORD IN THE AGE OF AI
AI is creating a new opportunity for startups to build systems of record rather than merely lightweight systems of action. AI agents require integrated, high-quality domain data, so businesses may replace entrenched software systems, increasing market turnover and expanding software vendors' potential from seat-based revenue to automated labor and outcomes.
2026.09.04
STARTUP ARR IS LESS SECURE THAN EVER, NEW RESEARCH SHOWS
Enterprise buyers are treating AI software very differently from traditional SaaS: 77% of companies surveyed by Madrona re-evaluate AI vendors every six months or continuously. That creates a “fast in, fast out” buying model where even landing an enterprise contract may no longer translate into durable recurring revenue.
2026.09.04
WHY DATA SOVEREIGNTY HAS BECOME A STRATEGIC IT PRIORITY
Data sovereignty is becoming less about simply knowing which region hosts your data and more about retaining control when vendors, regulations, infrastructure, or geopolitical conditions change. Portability, operational independence, and avoiding concentration risk should increasingly factor into architecture and vendor decisions.
2026.09.04
OKTA LAUNCHES AUTH0 TOOLS FOR AI AGENTS & B2B LOGINS
Okta has introduced new Auth0 capabilities for authenticating AI agents, agent-led commerce, and enterprise B2B identity. The launch also includes B2B Connect, which lets SaaS providers add enterprise SSO, federation, and multi-tenant identity capabilities without replacing their existing identity infrastructure.
2026.09.04
THE EVOLUTION OF JFROG AI CATALOG: YOUR AI CONTROL PLANE FOR AGENTIC DEVELOPMENT
JFrog says its AI Catalog has evolved from a secure model registry into an AI control plane for governing models, MCP servers, skills, plugins, and agent packages. The company presents semantic scanning, shadow AI detection, governed registries, and runtime Agent Guard enforcement as measures to prevent agents from consuming unapproved or malicious assets across the Agentic Development Lifecycle.
2026.09.04
CHATGPT, CLAUDE, AND GROK ALL WENT DOWN AT ONCE; ENTERPRISES NEED A BACKUP PLAN
ChatGPT, Claude, and Grok experienced overlapping multi-hour outages on Thursday, exposing a growing business-continuity problem as companies automate more workflows around AI services. IT teams may need to treat AI providers like other critical infrastructure, with fallback models, documented manual workflows, and outage planning.
2026.09.04
HOW NASA IS USING GRAPH TECHNOLOGY AND LLMS TO BUILD A PEOPLE KNOWLEDGE GRAPH
NASA's People Analytics team built a People Knowledge Graph with Memgraph, large language models, and secure AWS infrastructure to connect employees, skills, projects, and organizational data. The system supports expert discovery, project similarity analysis, workforce insights, and a GraphRAG chatbot, while remaining under development as NASA works to improve data quality, automation, and scale.
2026.09.04
WHEN AI'S HUMAN IN THE LOOP REALLY ISN'T
Many organizations claim to have human oversight of AI when reviewers can observe or flag decisions but cannot actually stop or override them. Effective human-in-the-loop controls require the right expertise, context, and authority, and need to be tested like any other operational control.
2026.09.04
ROOTLY AUTOMATICALLY SURFACES RELATED PAST INCIDENTS IN SLACK
Rootly's agent can now automatically post up to three relevant historical incidents into an active Slack incident channel, including why they match, severity, resolution time, and links to the previous incident.
2026.09.04
MICROSOFT TEAMS IS MOVING TO TEAMS.CLOUD.MICROSOFT
Microsoft is redirecting Teams web users to teams.cloud.microsoft, with admins advised to ensure the new domain is allowed through proxies, firewalls, and other network controls.
2026.09.04
Sponsored
THIS AD GOT YOUR ATTENTION. IMAGINE IT IN FRONT OF YOUR AUDIENCE.
8 million tech professionals read TLDR to stay on top of the news, ideas, and tools shaping their work. Put your brand in front of them. Become a TLDR sponsor [4].
2026.09.04
US COURT RULES GOOGLE WILL NOT HAVE TO SELL AD EXCHANGE AFTER LOSING ANTITRUST CASE
Google's ad tech dominance will largely remain intact despite the DOJ proving that it illegally restricted competition in parts of the online ad market. The court rejected the government's push to force Google to sell its ad exchange, avoiding a remedy that could have disrupted its broader advertising business. Google will likely face fines and business practice changes, but the exact penalties remain sealed for now. The outcome shows that even after three major antitrust losses, Google retains substantial control over digital advertising and can continue investing aggressively in emerging markets such as AI.
2026.09.04
‘A CRISIS OF AUTHENTICITY': THE TRUTH ABOUT FOOD INFLUENCERS
Positive coverage in food influencing is less a preference than an economic default, since restaurants that stop offering free meals or payment quickly see coverage dry up. That system still drives real reach, with Instagram Reels revenue growing from about $2 billion in 2022 to a projected $50 billion and roughly half of Instagram's time now spent on them. Most restaurant influencers still don't get paid well, with typical per-post rates staying in the low hundreds to low thousands and Instagram not paying for views at all. Viewers have noticed the pattern, and experts predict today's relentlessly positive accounts will lose ground to more critical voices.
2026.09.04
WIN REPORT: HOW SUPPLY-SIDE CRO LIFTED SIGN-UPS BY 15%
Supply-side CRO can unlock major marketplace growth by optimizing for the actual intent of people who provide capacity, not just the customers who buy. A tutoring marketplace found that many tutor recruitment visitors came from job search sites, so it replaced a generic registration form with a real student job request, clear job details, pay information, and “apply” language. The change increased tutor applications by 15%, or about 20,000 additional tutors in the first year. Supply-side improvements may not show immediate revenue, but they can strengthen marketplace capacity and create an overlooked source of high-impact conversion gains.
2026.09.04
THE IMPACT OF CONTENT FRESHNESS ON AI CITATIONS
Content cited by AI assistants runs 25.7% fresher than what ranks organically on Google, with roughly half of all citations tracing to material updated in the last 13 weeks. That edge decays fast, with a median citation half-life of about 4.5 weeks across platforms. Retrieval systems re-rank the entire candidate pool on every query rather than relying on a persistent ranking, so recency becomes an early filter for weeding out sources that might already be wrong. Perplexity and Grok weigh it most heavily, while Google's AI Overviews stay tied to core search and actually skew slightly older.
2026.09.04
CAPTURE ATTENTION THAT LASTS WITH 15-SECOND ENGAGED VIDEO VIEWS
Reddit launched a 15-second Engaged Video Views goal that optimizes video ads toward users likely to watch for at least 15 seconds or to completion. It's designed for longer-form creative such as product demos and testimonials, with advertisers paying only for qualifying views. Alpha testers saw average gains of 51% in 15-second view-through rate and 71% in completion rate. The feature is now available globally in public beta, with split testing available to compare it against 6-second views.
2026.09.04
BUILD VS. BUY: MARKETING TOOLS IN THE VIBE CODE ERA
AI is changing how marketing teams should think about build vs. buy. Teams should build when customization or proprietary data creates an advantage, and buy when vendors handle complex infrastructure that's expensive to maintain. They also warn against letting vibe-coded tools become one-off projects that only one person uses. When evaluating software, ask what would break if you canceled it and whether you're buying the product or simply a way to organize data you already own.
2026.09.04
CREATORS AS A SALES ENGINE: HOW CHINA IS RETHINKING THE ENTIRE CUSTOMER JOURNEY
China is showing how creators are becoming part of the sales infrastructure, not just a communications channel. On platforms like Rednote, creator content can intercept search, shape product research, trigger retargeting, and drive purchases through livestreams and in-platform ecommerce. The result is a more connected customer journey where content, search, community, and commerce work together rather than as separate channels. As platforms like TikTok expand into commerce, this model offers a preview of how social could take on more of the customer journey in Western markets.
2026.09.04
STICKI ROLLS IS TURNING STICKERS INTO THE NEW SOCIAL CURRENCY
Sticki Rolls turns stickers into social currency by designing the product around sharing and self-expression. It has expanded into bracelets, sticker books, pendants, and bag charms that fans can collect and show off. The brand also works with creators who shaped its original collections and uses unexpected spaces like nail salons to bring the product into beauty and fashion. Its pop-ups further turn fans into participants, with activities like sticker manicures and BFF bracelet-making.
2026.09.04
SKIMS MARKETING STRATEGY: KIM KARDASHIAN'S BILLION-DOLLAR BRAND
Behind SKIMS' celebrity headlines is a disciplined owned-channel operation, with the brand sending roughly five emails a week announcing drops, restocks, and retail collaborations. Before a major collection ships, SKIMS also seeds unboxing packages to thousands of micro-influencers across a range of body types and skin tones, producing unedited fit videos that build trust faster than a polished ad could. Both tactics are cheap, repeatable, and aimed at an audience that already opted in, the opposite of the celebrity-cash-grab reputation the brand is often given credit for.
2026.09.04