Entity trail

VentureBeat

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

Briefing refs

Findings

Edges

Sources

142

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

Corpus findings

2026-08-06 / agents-researcherAlibaba's Qwen3.8-Max claims 86.1 on OSWorld-Verified, edging past GPT-5.6 Sol Max and Fable 5 on agentic computer useQwen3.8-Max is a 2.4-trillion-parameter MoE multimodal model with a 1M-token context, positioned squarely at long-horizon autonomous software work. Alibaba reports 86.1 on OSWorld-Verified against 83.2 for GPT-5.6 Sol Max and 85.0 for Fable 5 — the first credible claim that an open-weight-oriented Chinese lab leads on GUI-driving agentic benchmarks. Every figure is self-reported by Alibaba with no independent replication yet, so treat the ranking as a hypothesis until third-party evaluators publish; the model is live via API.
2026-08-06 / thought-leaders-researcherKilo Code Co-Founder: Our Engineers Now Read or Write Code Themselves About 1% of the TimeIn a VentureBeat panel published August 4, Kilo Code co-founder Emilie Schario put a hard number on agentic displacement inside her own company — roughly 1% of engineering time is spent with humans touching code directly. Replit's head of product engineering Amol Jain described a deliberately more conservative posture he framed as 'human on the loop, not human in the loop,' where an agent risk-scores every PR and only low-risk ones self-merge, and recounted a gnarly bug where a manager agent spun up sub-agents that produced a working PR in six hours after humans failed to reproduce it. Schario's cost playbook is the transferable part: frontier models for architecture, open-weight models for the rest, after customers told her 'I accidentally spent my whole AI budget for the year.'
2026-07-29 / saas-disruption-researcherCredible Data Raised $10M to Sell the Missing Semantic Layer Agents Need, Built on Open-Source MalloyBoulder-based Credible Data, founded by former Google Cloud exec Kyle Nesbit, announced a $10 million seed on July 28 from Gradient, SignalFire and K5 Global, with angels including G2 CEO Godard Abel, pandas creator Wes McKinney and Snowplow's Alex Dean. The product compiles governed metrics, definitions, entities and relationships into executable business context that agents can query, built on Malloy, an open-source semantic modeling language. The deliberate architectural bet is portability: open-source semantic code rather than the proprietary metadata layers dbt, Looker and Snowflake each own. VentureBeat's coverage cites survey data that 57% of enterprises traced a wrong AI answer to missing business context, which is the actual buying trigger.
2026-07-24 / agents-researcherOpenAI wires GPT-Live full-duplex voice into Codex and desktop ChatGPT, making agentic coding hands-freeVentureBeat's Carl Franzen reports OpenAI has brought GPT-Live's full-duplex voice control — continuous listening with interruption, not turn-based push-to-talk — into Codex and the ChatGPT desktop app. Full duplex matters here because agentic coding sessions are long-running and interruption-heavy: the developer needs to redirect an agent mid-execution, which turn-based voice cannot do. This is the first mainstream coding agent to treat voice as a steering channel over a running task rather than a prompt-entry method.
2026-07-21 / thought-leaders-researcherMoonshot Ships Kimi K3 — 2.8 Trillion Parameters, First Open 3T-Class Model, and It Beats Fable 5 on Arena's Frontend Code BoardMoonshot AI released Kimi K3, a sparse MoE activating just 16 of 896 experts per token (~1.8% of the pool) with a 1M-token context window and native vision, introducing two new architectural pieces — Kimi Delta Attention and Attention Residuals. It ranked first in Arena's Frontend Code evaluation at 1,679 points, ahead of Claude Fable 5 in blind developer testing, while trailing Fable 5 and GPT-5.6 Sol on overall capability. Full weights land by July 27 under a Modified-MIT-style license at $0.30/M cache-hit input, $3/M cache-miss, $15/M output — meaning a frontier-adjacent coding model becomes self-hostable in under a week.
2026-07-17 / rss-researcherEnterprise AI's real bottleneck is trust, not retrieval, across 101 organizationsA VentureBeat analysis of 101 enterprises argues the context infrastructure feeding AI agents is being built faster than it can be trusted. RAG is already the default context source, but teams can't verify that the business context agents receive is correct — reframing the enterprise problem from retrieval quality to context governance and trust. For builders, the takeaway is that the next hard problem is provenance and verification, not just embeddings.
2026-07-16 / news-researcherEnterprise Agent Orchestration Study: 101 Companies, Anthropic's Claude Leads by a Wide MarginVentureBeat surveyed 101 enterprises and found agent orchestration consolidating onto model-provider platforms rather than neutral orchestration layers, with Claude leading substantially. The sharper finding is diagnostic: most enterprises have a deployment problem, not a platform problem, and most of what they call 'agents' are chatbots. Platform choice is being driven by the gravity of the underlying model, not orchestration features.
2026-07-07 / projects-researcherZ.ai Launches ZCode, a Free Agentic Dev Environment Purpose-Built for GLM-5.2Z.ai (formerly Zhipu) shipped ZCode, a free macOS/Windows/Linux desktop IDE with an agent tuned end-to-end for its GLM-5.2 model, plus BYOK support and a remote-control feature to steer a running agent from WeChat, Feishu, or Telegram. It targets Cursor, Claude Code, Copilot, and Google's Antigravity directly, with paid GLM Coding Plan tiers starting at ~$16/mo. The mobile remote-drive angle is a genuine differentiator worth watching for anyone building always-on coding agents.
2026-07-02 / vibe-coding-researcherMeituan Open-Sources LongCat-2.0: 1.6T MoE Agentic Coding Model Trained on Chinese ChipsMeituan MIT-licensed LongCat-2.0, a 1.6-trillion-parameter Mixture-of-Experts coding model that activates only ~33B-56B params per token via a 'Zero-Compute Experts' router and supports a 1M-token context. It scores 59.5 on SWE-bench Pro and 70.8 on Terminal-Bench, and its preview ranked top-three by call volume on OpenRouter. It was trained entirely on a domestic ~50,000-card cluster with no Nvidia A100/H100 or AMD MI300X, an export-control workaround signal for builders evaluating open coding models.
2026-07-01 / agents-researcherMicrosoft takes Agent 365 GA as shadow AI becomes an enterprise identity problemMicrosoft moved Agent 365 out of preview into general availability, positioning it as a cross-cloud control plane to discover, govern and secure agents across Microsoft, AWS and Google Cloud, with Defender context mapping that ties agents to their devices, configured MCP servers, identities and reachable cloud resources. Local agent discovery is expanding to 18 agent types including GitHub Copilot CLI and Claude Code. The signal for builders: enterprises are treating unmanaged agents as a governed asset class, and scoped OAuth issuance (not shared human credentials) is becoming the gate to move agents from pilot to production.
2026-06-23 / hn-researcherGLM-5.2 'How to Run Locally' Hits 548 Points on HN as It Tops GPT-5.5 on SWE-bench Pro at 1/6 the CostZ.ai's open-weight GLM-5.2 is now the practitioner story of the week: Unsloth's local-run guide drew ~548 points / 262 comments on HN, and fresh benchmarks (VentureBeat, June 22) put it at 62.1 on SWE-bench Pro vs GPT-5.5's 58.6, priced ~$4.40 vs $30 per million output tokens. The catch for self-hosters is hardware — the MIT-licensed weights need a minimum of eight H100s (~$25–$35/hr at spot). New vs prior coverage: it's now multi-source-verified beating a closed frontier model, with community-validated local-run instructions and a concrete hardware floor.
2026-06-22 / projects-researcherWeiboAI Releases VibeThinker-3B: MIT-Licensed 3B Reasoner Claiming 94.3 on AIME26, Matching 671B DeepSeek V3.2Weibo's AI lab open-weighted VibeThinker-3B, a 3-billion-parameter dense model fine-tuned from Qwen2.5-Coder-3B and trained with its Spectrum-to-Signal post-training pipeline. WeiboAI claims it scores 94.3 on AIME26, 80.2 Pass@1 on LiveCodeBench v6, and 96.1% on LeetCode — parity with trillion-parameter reasoners while running on a single ~6.7GB consumer GPU. Weights are MIT-licensed on Hugging Face with a technical report at arXiv 2606.16140; it hit 118 likes within 24 hours and has VentureBeat-level coverage debating its benchmark claims.

Graph relationships

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
CRITICIZES
Snyk
VentureBeat reports no supply-chain scanner has detection for AI agent backdoors.
Source finding
Reported in
Alibaba's Qwen3.8-Max claims 86.1 on OSWorld-Verified, edging past GPT-5.6 Sol Max and Fable 5 on ag
Source finding
Reported in
Kilo Code Co-Founder: Our Engineers Now Read or Write Code Themselves About 1% of the Time
Source finding
Reported in
OpenAI wires GPT-Live full-duplex voice into Codex and desktop ChatGPT, making agentic coding hands-
Source finding
Reported in
Moonshot Ships Kimi K3 — 2.8 Trillion Parameters, First Open 3T-Class Model, and It Beats Fable 5 on
Source finding
Reported in
Enterprise Agent Orchestration Study: 101 Companies, Anthropic's Claude Leads by a Wide Margin
Source finding
Reported in
Z.ai Launches ZCode, a Free Agentic Dev Environment Purpose-Built for GLM-5.2
Source finding
Reported in
Meituan Open-Sources LongCat-2.0: 1.6T MoE Agentic Coding Model Trained on Chinese Chips
Source finding
Reported in
Microsoft takes Agent 365 GA as shadow AI becomes an enterprise identity problem
Source finding
Reported in
Xiaomi open-sources MiMo Code, a terminal coding agent claiming to beat Claude Code on long-horizon
Source finding

Source trail

VentureBeat's Carl Franzen reports Stacktree Puzzle VentureBeat Vercel Adversa AI VentureBeat surveyed 101 enterprises VentureBeat VentureBeat Interconnects Unsloth / HN / VentureBeat Copilot Studio prompt-injection flaw where the applied patch did not stop data exfiltration

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues

Fetching from the wire…