This is the news. Agent dossiers — when to pick one, when to skip it — are on the agents page. Practical guides support your own testing and deployment decisions.
Anthropic filed a confidential S-1 with the SEC last week at a $900B-$1.2T implied valuation. The math does not close on software-revenue multiples — it closes on compute-asset multiples. Here is why your inference stack now depends on the oil pipeline, not the API.
Bloomberg reported on July 16 that Google scrapped its near-ready Gemini 3.5 Pro and restarted training after coding results came in worse than expected. The flagship Pichai promised for June is now a rebuild with no public release date. Here is what that actually means for anyone building on Google AI.
Six AI model launches in seven days. Five vendors. Four different pricing strategies. Kimi K3, DeepSeek V4, Grok 4.5, and Leanstral 1.5 just ended the single-vendor era. Here is the production router stack I would build on Monday if my inference bill were north of $50K a month.
Moonshot AI paused new Kimi K3 signups on July 19–20 after 48 hours of demand pushed its GPU fleet to the limit. Here's why a frontier open-weights model hitting a compute ceiling 4 days after launch complicates the open vs closed argument.
DeepSeek V4 officially went GA on July 13, 2026 — and with it came a pricing structure that every production API user needs to understand: peak hours (Beijing 9–12am and 2–6pm) now cost 2x off-peak rates. Here's what this means for your inference budget and why the off-peak math is genuinely absurd.
OpenAI's July 9 GPT-5.6 release isn't a single flagship — it's Sol, Terra, and Luna with a new multi-agent Ultra mode, programmatic tool calling, and pricing that makes Claude Fable 5 look expensive. Here's what practitioners actually need to know.
GPT-5.6 Sol, Terra, and Luna went GA on July 9, 2026 — two weeks after OpenAI held the models back at the U.S. government's request because they were too good at offensive cyber. The model is real. The precedent is worse.
OpenAI shipped GPT-5.6 Sol, Terra, and Luna to general availability on July 9, 2026. The coding benchmarks got the headlines. The cybersecurity numbers — ExploitBench2 47.9% to 73.5%, SEC-Bench Pro 45.8% to 71.2%, ExploitGym3 approximately 2x pass rate — are what actually move the needle, and the new Daybreak trusted-access regime is the most consequential deployment model any frontier lab has ever shipped.
A fintech I work with burned $112,000 in three weeks after flipping their support agent to o3-pro and not instrumenting reasoning tokens. Reasoning models broke every assumption you had about LLM cost predictability. Here is the OTEL-native, vendor-portable stack that fixes it.
Claude Code will happily `rm -rf` the wrong directory. Wire a 20-line `PreToolUse` hook that vets every Bash command against a denylist of foot-guns and exits non-zero to veto the dangerous ones. Twenty lines, no seatbelt excuses.
Every LLM gateway I've shipped in 18 months has been a glorified routing proxy with a config file. Portkey is the first open-source one that figured out the gateway layer is supposed to be a control plane, not a proxy.
SWE-Bench, tau-Bench, GAIA, OSWorld, WebArena — every public agent leaderboard in 2026 is rigged carnival theatre. Labs know it. Your CFO does not. The score that matters is the one you run yourself on your own tickets.
Mistral shipped Leanstral 1.5 — an Apache-2.0 open-weights MoE (119B total, 6.5B active) that solved 587 of 672 Putnam problems for ~$4 each, against Seed-Prover 1.5's $300+. Formal verification just became cheap.
Pre-counting tokens stops the obvious cost incidents. It does not stop the agent loop that spins for two hours calling gpt-4o-mini 40,000 times because a JSON schema validator keeps returning 400. You need a hard kill switch — a process-level budget that aborts mid-stream when the meter passes the cap. Here is the build, 30 lines.
Every 'Prompt Engineer — $250K + equity' LinkedIn post is a self-own. The job description is shallow, the skill ceiling is depressing, the vendors are writing the role to shift blame off their models, and the tooling that replaces hand-tuned prompting is shipping every six weeks. Hire an AI engineer. Retire the title.
SpaceXAI shipped Grok 4.5 yesterday at $2 input / $6 output per million tokens — roughly 10x cheaper than Opus 4.8 per solved SWE-Bench task once you count the 4.2x token reduction. The real moat is the Cursor data flywheel, not the leaderboard.
Most LLM serving stacks bolt structured generation on top of a chat-completion engine. SGLang flips the script — a frontend DSL and runtime co-designed for branching, parallel, and constrained LLM programs. If your workload looks like an agent, a tree-of-thought, or a JSON-schema-heavy pipeline, it deserves a serious look.
Stand up a Postgres + pgvector backend that indexes millions of embeddings and answers nearest-neighbor queries in milliseconds — with the exact schema, index choice, and a working query loop.
Your fancy multi-agent orchestration is one LLM pretending to be a team. Stop shipping it. Stop selling it. Stop pretending the trench coat is full.
Anthropic shipped Claude Sonnet 5 on June 30 with near-Opus-4.8 agentic performance at one-fifth the price, and quietly rug-pulled every team still routing serious work to Opus. The tokenizer change is a 30% price hike nobody is talking about.
The Model Context Protocol went from Anthropic experiment to industry standard in six months. Here is what it actually gets right, where production implementations are still a mess, and what you should be doing today.
Five frameworks I deleted in 2026 are good software. The sixth is 137 lines of Python handling 11M turns a week. Here is exactly when to use which runtime — and the 20-line agent.py you should have shipped instead of your agent platform.
The demos are real. The production systems are mostly theater. Here is the hard technical map of what computer use agents can actually do in July 2026, the three architectural approaches, the six failure modes that kill deployments, and the production stack with code that works.
The skills economy is the new software supply chain. 180,000+ skills on the public registries, 1M+ in private use, growing 4,200/week on the public side and 25,000/week across enterprise marketplaces. The SKILL.md format is the universal agent skill artifact, the install path is one command, the four marketplaces that matter are Skills.sh, the Claude Skills Registry, Hugging Face Skills Hub, and Microsoft Copilot Studio. Here is the format, the install mechanics, the security model, the economics ($380M direct, $1.5B annualized), and the stack I'd ship Monday morning if I were a team building a real skill product this quarter.