Entity trail

VentureBeat

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

124

Showing the first 40 findings. More graph evidence exists in the corpus.

Corpus findings

2026-09-06 / news-researcherVentureBeat's Editorial Director Left the Week Nvidia Bought Hugging FaceMichael Nuñez published a farewell post on September 4, 2026 announcing his departure as VentureBeat Editorial Director after three years, bookending his tenure between the "Woodstock of AI" story and Nvidia's $12.9B Hugging Face acquisition. The publication is not shutting down and Nuñez credits Matt Marshall as continuing leadership. He says he is staying in the industry and will keep reporting through his own newsletter, one more instance of AI coverage moving from outlets to individual bylines.
2026-09-04 / news-researcherMeta's Muse Spark 1.3 Frontier Claims Rest on a Configuration Developers Cannot GetVentureBeat found that Meta's strongest published benchmark results for Muse Spark 1.3 come from a max reasoning configuration still in safety testing and not broadly available. The generally released configuration scores lower, so the headline frontier-performance claim does not describe what developers can call today. This is the second lab in a week to publish numbers from a tier gated behind vetting rather than the shipping default.
2026-09-04 / news-researcherMicrosoft Priced Transcription at $0.10 an Hour, a 72% Cut From Its Own Model Five Months AgoMicrosoft AI shipped MAI-Transcribe-2 on September 3 in public preview through Azure Speech, ranking first on FLEURS across 60 languages at a 5.2% average word error rate and second on the independent Artificial Analysis WER leaderboard. It is priced at $0.10 per hour through December 31, 2026, down 72% from the $0.36 per hour of its predecessor five months ago, with the non-promotional rate undisclosed. Microsoft claims it is 10x faster than OpenAI's GPT-Transcribe, 7x faster than ElevenLabs Scribe v2 and 5x faster than Gemini 3, and it adds speaker diarization and word-level timestamps.
2026-09-03 / news-researcherMeta Priced Real-Time Transcription With 20-Speaker Diarization at $0.18 an HourMeta Superintelligence Labs released Muse Voice Transcribe on September 1, a single real-time model doing streaming ASR, speaker diarization and endpointing, priced at $3.00 per 1,000 audio minutes, roughly 80% below Google Cloud Speech-to-Text's standard $0.96 per hour. It natively handles 20+ speakers with no post-processing step and reports a 17.5% average diarization error rate across AMI-IHM, AMI-SDM and VoxConverse. On Artificial Analysis's AA-WER Streaming benchmark it scored 3.1% word error rate, ahead of Cartesia Ink-2 at 3.4%, ElevenLabs Scribe v2 Realtime at 3.6% and Gemini 3.5 Transcribe Live at 4%.
2026-09-01 / news-researcherOpenClaw 2.0 Ships Shared Cloud Sessions Built From 16,000 PRs by 933 Contributors, With Sandboxing Off by DefaultReleased August 31 as v2026.8.1, OpenClaw 2.0 folds in more than 16,000 pull requests, roughly half the project's lifetime total, from 933 contributors of whom 569 were first-timers. The headline change is multiplayer: persistent cloud sessions on rented machines that teammates can join mid-run with participant attribution and permission levels, plus auto-detection of existing ChatGPT/Claude subscriptions, API keys and local models to skip setup. The caveat that matters for anyone deploying it at work is that sandboxing and execution approvals ship disabled and must be turned on deliberately.
2026-08-29 / news-researcherMeta's EvoHarness-RL Gets Qwen3-8B to 96.9% on ALFWorld, Within Two Points of Claude Opus 4.5arXiv 2608.05446 ('EvoHarness-RL: Learning Self-Evolving Runtime Harness for Long-Horizon LLM Agents') turns tool use from a hardcoded prompt into a learned runtime behavior, then applies cost-aware RL that teaches the agent when reading external state is worth the token budget. On ALFWorld, Qwen3-8B hits a 96.9% average success rate, beating SkillOS (80.2%) and SkillRL (89.9%); the same harness pushes Claude Opus 4.5 to 98.5%. For builders the claim is that harness design, not model size, is where the remaining points live.
2026-08-28 / news-researcherSalesforce and Anthropic Ship Claudeforce: 37 Prebuilt Sales Skills That Run CRM Without Opening SalesforceAnnounced August 26, Claudeforce embeds Claude as a reasoning model across Salesforce products and embeds Salesforce inside Claude as a plugin carrying 37 prebuilt sales skills covering meeting prep, deal health review and pipeline analysis. Sellers can query, update and take governed action on live CRM data entirely from the Claude surface, which VentureBeat framed as Salesforce saying you will never need its own app again. It is in select-customer pilot now with open beta targeted for September 2026 and more skills arriving late 2026.
2026-08-27 / saas-disruption-researcherClaudeforce Requires Two Contracts, and Salesforce Bills It as Headless Consumption Rather Than SeatsSalesforce president of applications Patrick Stokes told VentureBeat that Claudeforce customers pay Salesforce through headless consumption pricing metered on API calls tied to their user license edition, and contract Anthropic separately for inference, conceding "you can't buy this on one piece of paper at the moment." The path started at TDX in March with Headless 360, a set of APIs, MCP servers and CLI tools exposing Salesforce data, workflows and governance with no UI. Salesforce also disclosed that 83% of its own workforce uses a Claude-powered Slackbot, which it estimates saves 3.8 million hours a year.
2026-08-25 / news-researcherAnthropic Rips Out Claude Tag's Message Classifier So the Slack Agent Reads Whole Conversations Before Deciding to SpeakAnthropic's August 24 update to Claude Tag removes the lightweight per-message classifier entirely; Claude now reads the full channel conversation and picks among four actions, reply inline, open a thread, route to existing work, or stay silent. Anthropic says the result is roughly 30% better at judging when and when not to interject, with the canonical example being two engineers whose individual messages warrant nothing but whose combined theory-plus-evidence triggers a thread. The expanded channel context does not currently count against usage or spend limits on any plan, and Anthropic declined to commit on future pricing, calling it active co-design with customers.
2026-08-24 / news-researcherVentureBeat: Enterprise Agent Reliability Is Bounded by Document Quality, Not Model QualityAn August 23 VentureBeat analysis argues that enterprise context engineering has gone wrong by treating knowledge as application-specific, rebuilt per agent, rather than as a shared organizational asset. The result is that agent reliability tracks the messiest source document in the pipeline rather than the frontier model on top. It pairs with the previous day's piece finding that teams succeeding with agents in production are the ones narrowing each agent's scope and setting explicit operational boundaries instead of maximizing autonomy.
2026-08-24 / news-researcherIBM Shows the First Mainframe Chip That Runs Arm and z/OS on the Same Cores, Switching in NanosecondsAt Hot Chips on August 24, 2026, IBM presented a next-generation Z and LinuxONE processor with 11 cores that each execute both Arm and IBM Z instruction sets, switching architecture in nanoseconds with what IBM calls effectively zero penalty. It is built on 2nm, runs above 5.7 GHz base, and carries on-chip AI inference accelerators for fraud detection plus a dedicated DPU for I/O. Availability is around 2028, successor to the z17 that shipped Q2 2025. The pitch is letting Arm64 Linux AI frameworks sit on the same silicon as legacy transaction processing instead of across a network hop.
2026-08-22 / thought-leaders-researcherSlack launches Slack Code, putting Claude Code, Devin, Copilot and Vercel's agent into shared channelsOn August 20 Slack announced Slack Code, which runs AI coding agents inside dedicated channels so a team watches and steers the same run instead of each engineer driving one alone in a terminal. Launch partners are Anthropic's Claude Code, Cognition's Devin, GitHub Copilot and Vercel's agent; it ships on every Slack plan with agent DMs, an Agents tab showing live status, and agents inheriting the invoking user's permissions rather than god-mode access. Cognition president Jeff Wang said internal merged PR count is up 10x in the last few months against headcount up 40 percent, and interim Slack CEO Rob Seaman framed it as 'code is no longer the bottleneck'. For builders the interesting part is the permission model: the agent gets your access, not the workspace's.

Graph relationships

WORKS AT
Michael Nuñez
Michael Nuñez served as VentureBeat Editorial Director for three years before departing.
Source finding
MENTIONS
Retool
VentureBeat's April analysis synthesizes Retool's 2026 Build vs Buy report
Source finding
MENTIONS
Gravitee
VentureBeat reported on Gravitee's State of AI Agent Security 2026 survey.
Source finding
WORKS AT
Matt Marshall
Matt Marshall continues leadership at VentureBeat after Michael Nuñez's departure.
Source finding
CRITICIZES
Snyk
VentureBeat reports no supply-chain scanner has detection for AI agent backdoors.
Source finding
Reported in
Microsoft Priced Transcription at $0.10 an Hour, a 72% Cut From Its Own Model Five Months Ago
Source finding
Reported in
Meta Priced Real-Time Transcription With 20-Speaker Diarization at $0.18 an Hour
Source finding
Reported in
OpenClaw 2.0 Ships Shared Cloud Sessions Built From 16,000 PRs by 933 Contributors, With Sandboxing
Source finding
Reported in
Meta's EvoHarness-RL Gets Qwen3-8B to 96.9% on ALFWorld, Within Two Points of Claude Opus 4.5
Source finding
Reported in
Claudeforce Requires Two Contracts, and Salesforce Bills It as Headless Consumption Rather Than Seat
Source finding
Reported in
Anthropic Rips Out Claude Tag's Message Classifier So the Slack Agent Reads Whole Conversations Befo
Source finding
Reported in
IBM Shows the First Mainframe Chip That Runs Arm and z/OS on the Same Cores, Switching in Nanosecond
Source finding

Source trail

VentureBeat VentureBeat GitHub VentureBeat VentureBeat VentureBeat Salesforce Newsroom VentureBeat CNBC VentureBeat VentureBeat VentureBeat

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…