Daily news on LLMs, AI agents, and the tools around them.
-
AI News — July 28, 2026
Moonshot shipped Kimi K3's weights — 2.8 trillion parameters, 1.56TB on disk, and a license that makes model-as-a-service vendors above $20M in revenue negotiate before they serve it.
-
AI News — July 27, 2026
Kimi K3's weights are not out yet: Moonshot's own Hugging Face repo carries a release timer set to 15:00 UTC today, hours after outlets began treating the drop as done.
-
Weekly recap
AI News — Week of July 20–26, 2026
Claude Opus 5 landed at half Fable 5's price with a per-request effort dial, Gemini 3.6 Flash undercut its own workhorse tier, and Cursor shipped routing — the week cost became a setting rather than a model choice.
-
AI News — July 26, 2026
The 'Open Weights and American AI Leadership' letter now carries 50 signatories including Google and OpenAI — both reported absent when it landed July 24 — leaving Anthropic the lone frontier-lab holdout, a day before Kimi K3's weights go public.
-
AI News — July 25, 2026
Anthropic launched Claude Opus 5 — its new default Opus, priced at $5/$25 per million (unchanged from Opus 4.8 and half of Fable 5's rate) with a per-request low/medium/high effort dial that trades cost for capability, and Anthropic's numbers put it around Fable 5-level intelligence with a new state-of-the-art on agentic coding, reframing the flagship tier as a cheaper, tunable model rather than a pricier one.
-
AI News — July 24, 2026
OpenAI launched Presence, an enterprise platform for deploying governed voice and chat agents whose distinguishing feature is a Codex-driven improvement loop — the coding agent reviews production transcripts and proposes agent-behavior changes for staff to approve — and OpenAI says it already resolves 75% of calls on its own support line and cut human handoffs 15 points in 10 days, marking a shift from selling model access to selling a managed, self-improving agent system.
-
AI News — July 23, 2026
Google shipped Gemini 3.6 Flash and two siblings — 3.5 Flash-Lite and a governments-only 3.5 Flash Cyber — but still no 3.5 Pro: the new workhorse cuts output-token use ~17% (up to 65% on DeepSWE) at a lower per-token price while scoring higher on every internal eval, the first concrete sign Google's Gemini pipeline is shipping again after three missed 3.5 Pro targets.
-
AI News — July 22, 2026
Claude Code 2.1.216–2.1.217 put the first hard limits on runaway multi-agent fan-out — a default cap of 20 concurrent subagents, no nested subagents unless you opt in, and a `--max-budget-usd` ceiling that now actually halts background agents — alongside a run of worktree/symlink workspace-escape fixes, shifting the release focus from interpreting allow-rules to bounding what a single prompt can spawn and spend.
-
AI News — July 21, 2026
Anthropic finalized Fable 5's place in its subscriptions: as of July 20 the model is a permanent part of Max and Team Premium at 50% of plan limits, while Pro and Team Standard lose included access — dropping to a one-time $100 usage credit and then $10/$50-per-million metering — ending six weeks of rolling extensions with a two-tier split rather than another deadline.
-
AI News — July 20, 2026
Anthropic shipped Claude Code 2.1.214 with another round of permission-model fixes — closing an `Edit(src/**)` allow-rule that auto-approved writes to nested directories anywhere in the tree and a permission-check bypass in Windows PowerShell 5.1 — a second straight release window aimed at the approval layer rather than the model, on an otherwise quiet day dominated by watch-list dates.
-
Weekly recap
AI News — Week of July 13–19, 2026
Coding became the axis that reshuffled the field: Google's Gemini 3.5 Pro was confirmed months late for coding shortfalls (~$200B off Alphabet) while China's Moonshot took #1 on an external frontend-coding board with the largest open-weight model ever announced.
-
AI News — July 18, 2026
Moonshot's Kimi K3 arrived as the largest open-weight model ever announced — a 2.8-trillion-parameter (≈50B-active) MoE — and took #1 on the external Frontend Code Arena, edging Claude Fable 5 and GPT-5.6 Sol on a leaderboard the two US flagships had led, though its own benchmark table still trails both; the weights themselves don't drop until July 27.
-
AI News — July 17, 2026
Google's Gemini 3.5 Pro missed its July 17 target: a July 16 Bloomberg report that the flagship is months behind schedule — with coding capabilities short of internal expectations — sent Alphabet down ~4% and erased roughly $200B in market value, turning the watch-list slip the last several briefings tracked into a confirmed delay with no new date.
-
AI News — July 15, 2026
GitHub pulled its /security-review command into the Copilot app (July 14): the AI vulnerability scan that shipped in Copilot CLI last month now runs inside everyday Copilot coding workflows, in public preview for Free, Pro, Business, and Enterprise users — the one substantive coding-agent move on an otherwise model-quiet day.