Older posts.
CI/CD in 2026: Agents Opened 1M PRs in 5 Months, Bots Write 1 in 5 Reviews, and Most Agent PRs Get No Human Look
What the pipeline looks like when AI agents open the pull requests: GitHub's coding agent went GA on 25 September 2025 and opened over a million PRs in five months, Copilot code review passed 60 million reviews and one in five on GitHub, CodeRabbit has reviewed 13 million PRs on $88 million raised, and the first studies find most agent PRs get no human review attention. The gates that still hold: the assigner cannot approve, an extra approval for bot authors, path fences, signed commits, SLSA provenance, and a workflow YAML that gives agent PRs their own lane.
13 minOct 5, 2026Browser Agents in 2026: 90.2% on OSWorld-Verified, 20.6% on OSWorld 2.0, and Three Agent Browsers Already Shut Down
Computer-use agents passed the 72.36 percent human baseline on OSWorld-Verified this year (90.2 percent for the top framework, 86.0 for Claude Fable 5), then scored 20.6 percent on the 108-task OSWorld 2.0 that replaced it; Operator, Project Mariner and ChatGPT Atlas were all discontinued within 19 months; prompt-injection success in Anthropic's own tests fell from 23.6 to 11.2 percent and no vendor claims zero; Cloudflare sells HTTP 402 access to agents while 52 percent of crawling is for training; and hosting settled on one headless Chrome per microVM at $0.02 to $0.12 a browser-hour.
13 minOct 5, 2026AIOps in 2026: 47% on ITBench, 10% on Hard On-Call, $6.50 an Investigation, and Four Outages Where the Automation Was the Incident
The incident-response agents of 2026 against the evidence: Datadog Bits AI SRE GA at about 6.5 credits ($6.50) per investigation, PagerDuty's approval-gated SRE Agent, Grafana's six GA agent tools, Splunk's AI SRE, Elastic buying Deductive, Resolve at a $1B headline valuation on $4M ARR, Traversal's $48M; benchmarks from 13.8% on ITBench in 2025 to 47% in May 2026, 20.7% exact root cause on OpenRCA 2.0, 10% on hard ORCA-Bench and 40% hallucinated causes; what the AWS, Azure, Cloudflare and Google postmortems say about automation; and why a human should hold the button.
14 minOct 5, 2026AI SOC Agents in 2026: 98% Accuracy Claims, 23 to 34% on the Benchmark, and 0% of Teams Letting Them Act Alone
What the security-operations agents from Microsoft, Google, CrowdStrike, Palo Alto, SentinelOne, Torq and a billion-dollar startup cohort actually do in 2026 and what is measured: a median of 100 alerts a day and 28 percent never investigated, 75-minute mean investigations, Microsoft's $4-an-hour compute units and Google's token meter, CrowdStrike's 98 percent triage claim against Meta's 23 to 34 percent benchmark, Anthropic's and Google's reports of attackers running agents against defenders, and a Sigma rule and CI step that keep a human on the merge button.
14 minOct 5, 2026Agent Memory in 2026: Vendors Claim 92.5 on LoCoMo, a Plain Filesystem Scored 74, and Every Memory Product Lost to BM25
Agent memory systems in 2026, kept honest: Mem0's 66.9 on LoCoMo in its own paper versus 92.5 on its managed platform a year later, Zep's 75.1 rebuttal and the three bugs behind Mem0's 66.0 for Zep, Letta's 74.0 with only a filesystem, MemoryAgentBench putting every commercial system under 45 on retrieval against BM25 at 60.5, Hindsight's 89.6, first-party memory from ChatGPT, Claude and Gemini with dates, both labs' compaction APIs, what re-reading 115k tokens costs at cache-read prices, a Postgres and pgvector layer with supersession, and why memory poisoning is still unfixed.
13 minSep 9, 2026Zero-Knowledge in 2026: Ethereum Blocks Proven in 3 Seconds for a Third of a Cent, a 52-Bit Security Gap, and Google Wallet Proving Your Age
Where zero-knowledge proofs actually stand: the Ethereum Foundation's real-time proving definition and the measured numbers on ethproofs (from 16 minutes and $1.69 to 3 seconds and $0.003 a block on consumer GPUs), why the security certification is the hard part (a 52-bit gap between the proven and claimed soundness of the hash used), the zkVM bugs fuzzers found this year, the L2 landscape at $43 billion with zero Stage 2 rollups, and the non-crypto uses shipping now: Google Wallet's ZK age proofs, the EU wallet's December deadline, zkML and zkTLS. Plus a proof you can generate on a laptop.
14 minSep 9, 2026On-Device Inference in 2026: 160 Tokens a Second on a Laptop, a 2-Bit Model in Every iPhone, and the Bandwidth Arithmetic That Decides What Stays Local
What actually runs on the device now: Apple's 3B model trained at 2 bits and the WWDC26 framework that lets any model plug in, Gemini Nano inside AICore, Windows ML and the quiet replacement of Phi Silica, the NPU TOPS numbers and why they are the wrong metric, the small-model field from Gemma 4 E2B to Qwen3.5-2B to LFM2.5 with licences, measured decode speeds from Google's own tables and an independent phone benchmark, the runtimes (llama.cpp with multi-token prediction, LiteRT-LM, ExecuTorch 1.0, MLX), the memory-bandwidth arithmetic that predicts all of it, and the model-extraction attacks nobody has fixed. With commands to run one today.
14 minSep 9, 2026AI for Science in 2026: One Phase III Trial, Two Operational Weather Models, and Erdős Proofs Nobody Wants to Read
What has actually been validated outside the lab that made the claim: rentosertib's Phase III start, ECMWF and NOAA running AI forecasts operationally, Evo 2's viable phage genomes, the Erdős-problem scoreboard and its rediscovery problem, the GNoME and MatterGen materials that turned out not to be new, and the licence and pip-install status of every open model an engineer can run this week.
16 minSep 9, 2026AI Power in 2026: 485 TWh, a 0.3 GW Stargate, a 116 GW Turbine Backlog, and 1,800 MW That Dropped Off the Grid in Seconds
The electricity numbers behind AI, kept honest: IEA's 485 TWh for 2025 and 950 by 2030, LBNL's 4.7 percent of US power heading to 12, what a prompt actually costs (0.24 Wh at Google, 30 times more in reasoning mode), which gigawatt campuses are energised versus announced (Stargate Abilene at 0.3 of 1.2 GW), where the power comes from (a 116 GW gas-turbine backlog with 2031 slots, nuclear restarts in 2027, one SMR construction permit), what the grid operators are doing about 233 GW queues and 1,800 MW load-loss events, rack power from 132 kW to 600 kW, and a script to measure your own job's energy.
14 minSep 9, 2026Agent-to-Agent Protocols in 2026: A2A Reached 1.0, Three ACPs Died or Merged, and MCP Learned to Do Most of the Same Job
What Google's Agent2Agent protocol actually specifies at v1.0, what changed when it moved into the Agentic AI Foundation in August, which vendors have shipped it versus announced it, how MCP's July 2026 revision reproduced A2A's task lifecycle almost state for state, the first real CVE on an A2A surface, what the production data says about how rare multi-hop agent systems are, and a working server and client on the 1.1 SDK with the breaking changes listed.
14 minSep 8, 2026Open Weights in 2026: Chinese Labs Ship the Frontier, the Licenses Grew Clauses, and Brussels Started Enforcing
DeepSeek V4, Kimi K3, GLM-5.3, Qwen3.8 and Xiaomi's MiMo are the open frontier now; Meta went closed and came back; the licences quietly acquired revenue thresholds; the EU AI Office got its enforcement powers on 2 August. What I found when I read every licence file, with the Hugging Face API calls that check them.
16 minSep 8, 2026Robot Foundation Models in 2026: 500,000 Hours of Human Video, 50-Step Action Chunks, and a Winning Score of 0.26
What vision-language-action models actually do, from π0's flow-matching action head to Gemini Robotics 2 and GR00T N1.7; who shipped humanoids (Unitree 5,500, AgiBot 15,000) and who didn't (Tesla, zero); the teleoperation behind the demos; and the benchmark numbers that say a home robot is not close. With a LeRobot dataset I loaded this morning.
14 minSep 8, 2026Confidential Computing for AI in 2026: Apple Moved to Google's TEEs, a $1,000 Interposer Broke the Keys, and Firecracker Still Can't Do It
What a hardware enclave actually promises for an AI workload, what it costs (4 to 28% depending on what you measure), what broke it this year (TEE.fail, Battering RAM, WireTap), who runs inference in one (Apple, Google, Meta, a dozen startups) and who doesn't say (OpenAI, Anthropic). Plus the state of SEV-SNP, TDX and Arm CCA in the kernel, why my microVM platform cannot offer it yet, and the commands to verify an attestation yourself.
15 minSep 8, 2026AI Found the Bugs: 23,000 Findings in a Month, a 3-Day Patch Mandate, and the End of curl's Bounty
From AIxCC's 18 real bugs in August 2025 to Project Glasswing's 23,019 findings by May 2026 and the first AI-written zero-day exploited in the wild. What the cyber reasoning systems, Big Sleep, Codex Security and Mythos Preview actually did, why bug bounties are drowning, why CISA now wants federal patches in three days, and a workflow for a small team that keeps the humans looking only at validated findings.
13 minSep 8, 2026Agent Observability in 2026: The Spans Are Standard, the Standard Isn't Stable, and Nobody Reads the Logs
OpenTelemetry's GenAI conventions moved to their own repo in June and still carry the Development badge; every vendor from Datadog to Grafana ingests them anyway. I instrumented an agent loop with the official Python library, show the exact spans it emits, and go through what the 2026 eval papers say actually catches failures: pass^k, log inspection, and a step budget.
13 min