What changed in AI today

• Sergiy Dybskiy of Sentry and Ben Homes of Warp are presenting the sponsored workshop.
• The sponsor says Sentry integration enables automated debugging with full context and captures agent metrics for evals.
• Sessions cover what Warp factories are and how Warp uses Sentry internally.
• The source gives no date, price, or registration details for the workshop.

Why it matters for marketers: Content teams should treat these debugging and eval claims as sponsor-stated and unverified, and label any coverage as sponsored.

• MIT study by Neil Thompson's lab says AI may soon transform probability, statistics, economic theory, cosmology and quantum physics.
• Share of theory problems solved by top models rose from about 2 percent in mid-2024 to about 35 percent by mid-2026.
• Models tested were OpenAI's o1, o3, GPT-5 through GPT-5.5; they had no internet access and had to solve problems unaided.
• Researchers say experts will still be needed to verify solutions, since models can produce errors on complex problems.

Why it matters for marketers: Content teams covering science should expect rapid AI progress in theory-heavy fields, and fact-checking AI-generated technical claims will remain necessary.

• Muse Wonder costs $4 per million input tokens and $20 per million output tokens via the Meta Model API.
• Wonder Full is limited to Muse Code subscribers and is not served through the API.
• Meta says Wonder scores higher on MemoryBench than Fable, Opus, or Astra, per its own testing.
• Models were reportedly trained at Meta's Prometheus data center in central Ohio.

Why it matters for marketers: Content teams building long-running assistants may get a stronger memory option, but pricing is comparable to rival frontier models and Meta's benchmark claims are self-reported.

• Sub-agents or coworker agents get dedicated identities and a 4-layer memory, per the sponsor.
• Agents can orchestrate across Gemini, Claude, and open models, and connect to any MCP server.
• Governance runs through Agent Sandbox and Agent Gateway, with real-time spend caps for cost control.
• A Gemini at Work keynote is available to watch, and on-demand sessions cover building and governing agents.

Why it matters for marketers: Teams evaluating agent platforms can compare built-in policy enforcement and spend caps. The source gives no pricing or general availability details.

• Sponsored Algolia content says enterprise search is moving from returning links to carrying out actions.
• Algolia says its retrieval and ranking layer powers enterprise intent routers.
• The piece describes turning natural language requests into actions it calls safe and auditable.
• It also covers building an indexed tool catalog and ranking candidates under policy constraints.

Why it matters for marketers: Content teams covering enterprise search should label this as sponsored vendor claims, since the source gives no benchmarks, pricing, dates or availability.

• Step 5 Preview costs $1 per million input tokens and $2.70 per million output tokens, per the sponsored item.
• Step 5 Preview reportedly scores near Claude Opus 5 (Medium) on Artificial Analysis's Intelligence Index, with a 1M-token context.
• Step 5 Preview is on OpenRouter and works in OpenCode, Kilo Code, Cline, omp and Hermes Agent; more integrations are planned.
• StepAudio 3 ASR ranks first in Artificial Analysis non-streaming transcription at 1.7% WER; Realtime ranks first in two conversational categories.

Why it matters for marketers: Teams building coding agents or budgeting for voice features can compare StepFun's per-token prices and benchmark rankings against current options.

• Preview API is live on Mistral Studio; open weights are promised by the end of the month.
• Model has 1 trillion parameters, 52 billion active, and is natively multimodal, per Mistral.
• Scores 82% on a real-vulnerability reproduce-and-patch test and 93% on Cybench, which Mistral says is a top open-weight result.
• Trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral's own European datacenters.

Why it matters for marketers: Content and security teams gain an open-weight option with fewer provider-level refusals, but should wait for released weights and independent checks before relying on the self-reported benchmarks.

• Full weights under Apache 2.0 are due this month. The text-only model has 23B active parameters per token.
• Team posts cite 23.8T pretraining tokens. Reflection's claimed 80.9 on SWE-bench Verified is unverified, and a tech report is promised.
• Observers reportedly place Beam around GLM-5.2 level and below DeepSeek V4 Flash on some benchmarks.
• Artificial Analysis, which had early access, expects Beam to be among the most token-efficient open models for its intelligence.

Why it matters for marketers: Content and marketing teams citing open-model benchmarks should wait for the promised weights and tech report before repeating Beam's self-reported scores.