Technology
50 articles in this category
Most read in Technology, last 30 days
- 01
How to Use Claude AI Free: 14 Methods That Actually Work
- 02
Dow's Carbon Ledger Accelerates Sustainability
- 03
Snowflake Ventures enterprise AI funds trust layer
- 04
AI Generative Apps: Growth & Global Adoption
- 05
Pomerium's Nick Taylor Enhances OpenClaw Security and UX
- 06
Brave API Undercuts Google Maps Pricing
Cloudflare made Kitesurf open source for agents
Cloudflare open sourced Kitesurf, its Workers based browser for AI agents, under Apache 2.0.
Cloudflare says AI pentests now run themselves
Cloudflare says 50-plus AI pentest frameworks now scan customers autonomously, up 30x since Oct 1.
AI cut attacks that took months to 20 minutes
AI cut full exploit chains to 20 minutes, and Black October proved it on PHP servers with no login.
OpenAI wants AI to patch as fast as it finds
OpenAI demoed Daybreak Red Team and Defense Factory, with GPT6 Cyber results in Firefox, PHP and Linux spanning remote and local attacks.
Rust plus Arrow beats the Lambda detour
Eclipse Aerospace ditched Lambda-to-Avro and decodes flight data in Spark with Rust and Arrow.
Databricks bets on sovereign clouds in Europe
Databricks will add three sovereign builds to its 17 European regions across Azure and AWS.
CodeRabbit says its AI reviews block AI hacks
CodeRabbit CEO Harjot Gill says cyber models pen test its platform and code review is the bottleneck for coding agents.
Uber moved Presto to Velox, cut CPU by two thirds
Uber moved 90% of eligible Presto queries to Velox, cutting CPU by two thirds on 100 PB a day.
Isomorphic bets drug design leaves the lab
Isomorphic Labs president Max Jaderberg says AI must reach experimental accuracy to fix drug discovery cost and clinic failures.
Shultz Says Fear Silenced Theranos Staff
Tyler Shultz told Y Combinator his first day at Theranos showed no tests were validated and dissent meant threats.
Anthropic retools Claude Code for hands-off runs
Anthropic details how to run dozens of Claude Code sessions safely with auto mode, deny rules and sandboxes.
Brave will hide clicks to starve scrapers
Brave will route Search clicks through brave.com/go to block scrapers without tracking which result you picked.
Air Gaps Don't Have to Mean Dark Data
Bright Machines streams factory data out one auditable path with zero inbound access, queryable in minutes.
Databricks maps how stolen logins blend in
Databricks details six stages of attacks that use stolen tokens and how verbose logs expose them.
Your SKILL.md Teaches Little New
Intel's Nikolay Petrov says models already know 70% of popular SKILL.md content. Cut it.
AWS Chile Region opens with $4 billion investment
AWS opened its Chile Region to select customers with a $4 billion plan ahead of full availability.
GLP-1s Got Big by Making Patients Customers
Ro CEO says GLP-1s grew by making the patient the customer, cutting $1,300 list prices to $149 cash.
Linear Says Coding Agents Miss Product Context
Linear's Tom Moor showed triage and review agents that keep product intent with coding agents.
America stopped rewarding builders
a16z Blog says the Senate permitting bill would force deadlines, one review and final permits to rebuild US building speed.
Your Agent Isn't Lazy, Your CLI Is Noisy
Bhardwaj says pipes are bug reports, but most greps need better flags, not more analytics.
Self-Hosted LLMs Must Learn to Say No
Audible and SAP engineers tell AI Engineer that self-hosted LLMs need ingress admission control to survive 40-call agent bursts.
Dharmesh Shah says the SaaSpocalypse isn't happening
HubSpot co-founder Dharmesh Shah tells a16z AI aids builders, but offers no retention or agent numbers to price the bet.
OpenAI turns defense factory into Cloud product
OpenAI demoed Codex Security Cloud at DevDay 2026, turning its internal defense factory into a GitHub scanner.
Reading a skill file cannot prove it is safe
Zenity detonated thousands of agent skills, found 100 malicious ones, and showed static review misses remote payloads.
Resend drops approval gates, eats spam risk
Resend CEO Zeno Rocha says no-approval sending plus fast blocks and queue shaping keeps phishing off Gmail and Yahoo.
Agents need a shared runtime, Schroeder says
Justin Schroeder announced a standard agent runtime spec, arguing scoped agents beat broad bots on trust and results.
OpenAI turns Astra on itself to catch flaws
Greg Brockman said OpenAI ran Astra on its own systems, found critical flaws and now urges defenders to scan with frontier models first.
Your agent scolding is training data
Brian Douglas says every agent pushback is training data if teams capture and label traces.
Crusoe cuts GLM-5.3 startup to 31 seconds
Crusoe and NVIDIA cut GLM-5.3 startup from 8 minutes to 31 seconds by restoring GPU snapshots from local NVMe.
NVIDIA move to secure AI agents Kubernetes favors operators
NVIDIA donated OpenShell to CNCF Sandbox and added Kubernetes tools for AI factories, with early tests showing startup cuts from minutes to seconds.
Git patches six flaws with code execution risk
Git shipped 2.56.1 to fix six CVEs, including code execution via recursive clone and path traversal via protocol v2.
Ring's robots patrol, but proof is missing
Ring says mobile patrols beat fixed cameras, but FCC hold, Protect fees and late ship dates undercut the claim.
Amazon sets 12-day Black Friday clock
Amazon will run Black Friday Week and Cyber Monday Nov. 19-30 with double deal drops and early Haul and Autos windows.
Isaac Lab 3.0 trains robots for delicate assembly work
Nvidia introduced GR00T 2 and Isaac Lab 3.0 at CoRL to rehearse dexterous work in simulation before factory deployment.
Leo's Vulcan debut ends the slow rollout
Amazon Leo will launch 40 satellites on its first Vulcan flight, starting a 38-launch run aimed at Starlink scale.
Every agent framework wants a sandbox now
Daytona's Mislav Ivanda says every agent framework converged on create, run, delete for sandboxes. I am not convinced that settles security.
AI agents will fake a human to pass checks
Stanford tests show AI agents used voice and video clones to pass human checks in up to 65% of runs.
Your Inference Cache Already Sees Attacks
Ari Azarafrooz showed how KV cache anomalies can catch prompt injection and tool abuse without reading raw text.
Uber let trunk break to ship AI code faster
Uber's merge queue hit 130 requests an hour on AI diffs, so engineers let trunk break and fix after.
KetchUp bets wood shops skip the cloud
KetchUp ships a free, open-source 3D modeler for furniture with local files and SketchUp import.
YC Founder Says Cancer Is Now a Data Problem
AMA co-founder Alfredo told Y Combinator personalized AI labs can mine each tumor's genome to guide care now.
Cloudflare says agents break website trust
Cloudflare warns remotely hosted AI agents hide users from sites, breaking fraud controls and accountability.
Kepler says lookup must beat search
Vinoo Ganesh says finance AI must look up numbers by pointer, not generate them, after $270M chased verification.
The Hidden Capital Crunch Behind Silicon Valley’s AI Boom Recap: Every Talk and Launch Explained
This recap of The Hidden Capital Crunch Behind Silicon Valley’s AI Boom covers all 7 talks with speakers from Session and Abstract Ventures.
Virtual cells still promise more than they prove
Fabian Theis says virtual cells must steer experiments, but perturbation data and benchmarks lag the promise.
zstd pulls are faster, defaults are not
Crusoe says zstd cuts cold pulls up to 3x, but the win needs new digests and containerd 2.2 tuning.
Microsoft Is Pulling Back From Oracle AI Rental
Microsoft is winding down its Oracle AI capacity rental, Bloomberg reported, testing its own buildout.
Keycard and Temporal Replaced Static Agent Keys With Expiring Tokens
Kim Maida and Mason Egger replaced static API keys with Token Exchange and Temporal durable execution, cutting token lifetime to minutes.
Uber finds 5.5x Spark write gain in lab test
Uber says Onehouse Quanton cut Hudi write rounds up to 5.5x in synthetic tests, but gains are write-path only.
GitHub rebuilt Git for 7,200 pushes a second
GitHub demoed Project Rubicon, a new Git backend that handles 7,200 pushes per second and 5,000 concurrent agents on one repo.
