Scribe

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

Scribe AI

OpenClaw has 500,000 instances and no enterprise kill switch

“Your AI? It’s my AI now.” The line came from Etay Maor, VP of Threat Intelligence at Cato Networks, in an exclusive interview with VentureBeat at RSAC 2026 — and it describes exactly what happened to a U.K. CEO whose OpenClaw instance ended up for sale on BreachForums. Maor's argument is that the industry handed AI agents the kind of autonomy it would never extend to a human employee, discarding zero trust, least privilege, and assume-breach in the process. The proof arrived on BreachForums three weeks before Maor’s interview. On February 22, a threat actor using the handle “fluffyduck” posted a listing advertising root shell access to the CEO’s computer for $25,000 in Monero or Litecoin. The shell was not the selling point. The CEO’s OpenClaw AI personal assistant was. The buyer would get every conversation the CEO had with the AI, the company’s full production database, Telegram bot tokens, Trading 212 API keys, and personal details the CEO disclosed to the assistant about family and finances. The threat actor noted the CEO was actively interacting with OpenClaw in real time, making the listing a live intelligence feed rather than a static data dump. Cato CTRL senior security researcher Vitaly Simonovich documented the listing on February 25. The CEO’s OpenClaw instance stored everything in plain-text Markdown files under ~/.openclaw/workspace/ with no encryption at rest. The threat actor didn't need to exfiltrate anything; the CEO had already assembled it. When the security team discovered the breach, there was no native enterprise kill switch, no management console, and no way to inventory how many other instances were running across the organization. OpenClaw runs locally with direct access to the host machine’s file system, network connections, browser sessions, and installed applications. The coverage to date has tracked its velocity, but what it hasn't mapped is the threat surface. The four vendors who used RSAC 2026 to ship responses still haven't produced

Slack adds 30 AI features to Slackbot, its most ambitious update since the Salesforce acquisition

Slack today announced more than 30 new capabilities for Slackbot, its AI-powered personal agent, in what amounts to the most sweeping overhaul of the workplace messaging platform since Salesforce acquired it for $27.7 billion in 2021. The update transforms Slackbot from a simple conversational assistant into a full-spectrum enterprise agent that can take meeting notes across any video provider, operate outside the Slack application on users' desktops, execute tasks through third-party tools via the Model Context Protocol (MCP), and even serve as a lightweight CRM for small businesses — all without requiring users to install anything new. The announcement, timed to a keynote event that Salesforce CEO Marc Benioff is headlining Tuesday morning, arrives less than three months after Slackbot first became generally available on January 13 to Business+ and Enterprise+ subscribers. In that short window, Slack says the feature is on track to become the fastest-adopted product in Salesforce's 27-year history, with some employees at customer organizations reporting they save up to 90 minutes per day. Inside Salesforce itself, teams claim savings of up to 20 hours per week, translating to more than $6.4 million in estimated productivity value. "Slackbot is smart. It's pleasant, and I think it's endlessly useful," Rob Seaman, Slack's executive vice president and general manager, told VentureBeat in an exclusive interview ahead of the announcement. "The upper bound of use cases is effectively limitless for it." The release signals Slack's clearest bid yet to become what Seaman and the company's leadership describe as an "agentic operating system" — a single surface through which workers interact with AI agents, enterprise applications, and one another. It also marks a direct challenge to Microsoft, which has spent the past two years embedding its Copilot assistant across the entirety of its productivity stack. From simple chatbot to autonomous coworker: six new capabilities that

Cohere Transcribe: Speech Recognition

Softr launches AI-native platform to help nontechnical teams build business apps without code

Softr, the Berlin-based no-code platform used by more than one million builders and 7,000 organizations including Netflix, Google, and Stripe, today launched what it calls an AI-native platform — a bet that the explosive growth of AI-powered app creation tools has produced a market full of impressive demos but very little production-ready business software. The company's new AI Co-Builder lets non-technical users describe in plain language the software they need, and the platform generates a fully integrated system — database, user interface, permissions, and business logic included — connected and ready for real-world deployment immediately. The move marks a fundamental evolution for a company that spent five years building a no-code business before layering AI on top of what it describes as a proven infrastructure of constrained, pre-built building blocks. "Most AI app-builders stop at the shiny demo stage," Softr Co-Founder and CEO Mariam Hakobyan told VentureBeat in an exclusive interview ahead of the launch. "A lot of the time, people generate calculators, landing pages, and websites — and there are a huge number of use cases for those. But there is no actual business application builder, which has completely different needs." The announcement arrives at a moment when the AI app-building market finds itself at an inflection point. A wave of so-called "vibe coding" platforms — tools like Lovable, Bolt, and Replit that generate application code from natural language prompts — have captured developer mindshare and venture capital over the past 18 months. But Hakobyan argues those tools fundamentally misserve the audience Softr is chasing: the estimated billions of non-technical business users inside companies who need custom operational software but lack the skills to maintain AI-generated code when it inevitably breaks. Why AI-generated app prototypes keep failing when real business data is involved The core tension Softr is trying to resolve is one that has plag

Nvidia-backed ThinkLabs AI raises $28 million to tackle a growing power grid crunch

ThinkLabs AI, a startup building artificial intelligence models that simulate the behavior of the electric grid, announced today that it has closed a $28 million Series A financing round led by Energy Impact Partners (EIP), one of the largest energy transition investment firms in the world. Nvidia’s venture capital arm NVentures and Edison International, the parent company of Southern California Edison, also participated in the round. The funding marks a significant escalation in the race to apply AI not just to software and content generation, but to the physical infrastructure that powers modern life. While most AI investment headlines have centered on large language models and generative tools, ThinkLabs is pursuing a different and arguably more consequential application: using physics-informed AI to model the behavior of electrical grids in real time, compressing engineering studies that once took weeks or months into minutes. "We are dead focused on the grid," ThinkLabs CEO Josh Wong told VentureBeat in an exclusive interview ahead of the announcement. "We do AI models to model the grid, specifically transmission and distribution power flow related modeling. We can calculate things like interconnection of large loads — like data centers or electric vehicle charging — and understand the impact they have on the grid." The round drew participation from a deep bench of returning investors, including GE Vernova, Powerhouse Ventures, Active Impact Investments, Blackhorn Ventures, and Amplify Capital, along with an unnamed large North American investor-owned utility. The company initially set out to raise less than $28 million, according to Wong, but strong demand from strategic partners pushed the round higher. "This was way oversubscribed," Wong said. "We attracted the right ecosystem partners and the right capital partners to grow with, and that's how we ended up at $28 million." Why surging electricity demand is breaking the grid's legacy planning tools The timing

Cohere's open-weight ASR model hits 5.4% word error rate — low enough to replace speech APIs in production pipelines

Enterprises building voice-enabled workflows have had limited options for production-grade transcription: closed APIs with data residency risks, or open models that trade accuracy for deployability. Cohere's new open-weight ASR model, Transcribe, is built to compete on all four key differentiators — contextual accuracy, latency, control and cost. Cohere says that Transcribe outperforms current leaders on accuracy — and unlike closed APIs, it can run on an organization's own infrastructure. Cohere, which can be accessed via an API or in Cohere’s Model Vault as cohere-transcribe-03-2026, has 2 billion parameters and is licensed under Apache-2.0. The company said Transcribe has an average word error rate (WER) of just 5.42%, so it makes fewer mistakes than similar models. It’s trained on 14 languages: English, French, German, Italian, Spanish, Greek, Dutch, Polish, Portuguese, Chinese, Japanese, Korean, Vietnamese and Arabic. The company did not specify which Chinese dialect the model was trained on. Cohere said it trained the model “with a deliberate focus on minimizing WER, while keeping production readiness top-of-mind.” According to Cohere, the result is a model that enterprises can plug directly into voice-powered automations, transcription pipelines, and audio search workflows. Self-hosted transcription for production pipelines Until recently, enterprise transcription has been a trade-off — closed APIs offered accuracy but locked in data; open models offered control but lagged on performance. Unlike Whisper, which launched as a research model under MIT license, Transcribe is available for commercial use from release and can run on an organization's own local GPU infrastructure. Early users flagged the commercial-ready open-weight approach as meaningful for enterprise deployments. Organizations can bring Transcribe to their own local instances, since Cohere said the model has a more manageable inference footprint for local GPUs. The company said they were able to

z.ai debuts faster, cheaper GLM-5 Turbo model for agents and 'claws' — but it's not open-source

Chinese AI startup Z.ai, known for its powerful, open source GLM family of large language models (LLMs), has introduced GLM-5-Turbo, a new, proprietary variant of its open source GLM-5 model aimed at agent-driven workflows, with the company positioning it as a faster model tuned for OpenClaw-style tasks such as tool use, long-chain execution and persistent automation. It's available now through Z.ai's application programming interface (API) on third-party provider OpenRouter with roughly a 202.8K-token context window, 131.1K max output, and listed pricing of $0.96 per million input tokens and $3.20 per million output tokens. That makes it about $0.04 cheaper per total input and output cost (at 1 million tokens) than its predecessor, according to our calculations. Model Input Output Total Cost Source Grok 4.1 Fast $0.20 $0.50 $0.70 xAI Gemini 3 Flash $0.50 $3.00 $3.50 Google Kimi-K2.5 $0.60 $3.00 $3.60 Moonshot GLM-5-Turbo $0.96 $3.20 $4.16 OpenRouter GLM-5 $1.00 $3.20 $4.20 Z.ai Claude Haiku 4.5 $1.00 $5.00 $6.00 Anthropic Qwen3-Max $1.20 $6.00 $7.20 Alibaba Cloud Gemini 3 Pro $2.00 $12.00 $14.00 Google GPT-5.2 $1.75 $14.00 $15.75 OpenAI GPT-5.4 $2.50 $15.00 $17.50 OpenAI Claude Sonnet 4.5 $3.00 $15.00 $18.00 Anthropic Claude Opus 4.6 $5.00 $25.00 $30.00 Anthropic GPT-5.4 Pro $30.00 $180.00 $210.00 OpenAI Second, Z.ai is also adding the model to its GLM Coding subscription product, which is its packaged coding assistant service. That service has three tiers: Lite at $27 per quarter, Pro at $81 per quarter, and Max at $216 per quarter. Z.ai’s March 15 rollout note says Pro subscribers get GLM-5-Turbo in March, while Lite subscribers get the base GLM-5 in March and must wait until April for GLM-5-Turbo. The company is also taking early-access applications for enterprises via a Google Form, which suggests some users may get access ahead of that schedule depending on capacity. z.ai describes GLM-5-Turbo as designed for “fast inference” and “deeply optimized for real-wor

OpenAI’s Frontier puts AI agents in a fight SaaS can’t afford to lose

When OpenAI launched Frontier in February, the announcement was described as a platform for enterprise AI agents. What it actually signalled was a challenge to the revenue architecture underpinning the software industry. Frontier is designed to act as a semantic layer in an organisation’s existing systems, connecting data warehouses, CRM platforms, ticketing tools, and internal […] The post OpenAI’s Frontier puts AI agents in a fight SaaS can’t afford to lose appeared first on AI News.

Weekly Rules Review: 2026-03-09

## Weekly Rules Documentation Review - 2026-03-09 ### Overall Health Assessment The rules documentation is in **good shape overall**. All 11 rule files are relevant, and the vast majority of file paths and code patterns they reference still exist in the codebase. Two files need minor updates, and one has a moderate accuracy issue. --- ### Audit Results #### AGENTS.md **Status**: Keep **Reasoning**: Core agent guide is accurate and well-structured. The rules index table, project setup instructions, pre-commit checks, and general guidance are all current. References to `npm run ts`, `tsgo`, TanStack Router, and Base UI are correct. --- #### rules/electron-ipc.md **Status**: Keep **Reasoning**: High-value, comprehensive guide. All referenced file paths exist (`src/ipc/contracts/core.ts`, `src/ipc/types/*.ts`, `src/ipc/handlers/base.ts`, `src/lib/queryKeys.ts`). The `pendingStreamChatIds` pattern still exists in `useStreamChat.ts`. The `writeSettings` shallow merge warning is still relevant (confirmed by recent fix in commit ef4ec84 preventing stale settings reads). --- #### rules/local-agent-tools.md **Status**: Keep **Reasoning**: Concise and accurate. `modifiesState` flag is actively used across many tool files. `buildAgentToolSet` exists in `tool_definitions.ts`. `handleLocalAgentStream` exists in `local_agent_handler.ts` with `readOnly`/`planModeOnly` guards confirmed. `todo_persistence.ts` exists. `fs.promises` guidance remains relevant for Electron main process. --- #### rules/e2e-testing.md **Status**: Needs Update **Reasoning**: Mostly accurate and high-value, but has two inaccuracies in helper method references. **Issues Found**: - Lines 64-75: References `po.clearChatInput()` and `po.openChatHistoryMenu()` as methods on PageObject directly, but they actually live on the `chatActions` sub-component: `po.chatActions.clearChatInput()` and `po.chatActions.openChatHistoryMenu()`. This contradicts the sub-component pattern documented earlier in the same file (lines 29-43). **Suggested Changes**: - Update the Lexical editor section examples to use `po.chatActions.clearChatInput()` and `po.chatActions.openChatHistoryMenu()` instead of `po.clearChatInput()` and `po.openChatHistoryMenu()` --- #### rules/git-workflow.md **Status**: Keep **Reasoning**: Comprehensive and high-value. Contains many hard-won learnings about fork workflows, `gh pr create` edge cases, GitHub API workarounds, and rebase conflict resolution patterns. The `GITHUB_TOKEN` workflow chaining limitation and the `--input` pattern for special characters are particularly valuable. Recent commits (51fc07e - GitHub App tokens) confirm this area is actively evolving. **Note**: This is the longest rules file (123 lines). Some of the very specific rebase conflict tips (React component wrapper conflicts, refactoring conflicts) may be overly situational, but they're low-cost to keep. --- #### rules/base-ui-components.md **Status**: Keep **Reasoning**: All referenced component files exist (`context-menu.tsx`, `tooltip.tsx`, `accordion.tsx`). `@base-ui/react` is in the project dependencies. The TooltipTrigger `render` prop guidance and Accordion API differences from Radix are high-value patterns that prevent common mistakes. --- #### rules/database-drizzle.md **Status**: Keep **Reasoning**: Short and high-value. `drizzle/meta/_journal.json` exists. Migration conflict resolution guidance is important for rebase workflows. --- #### rules/typescript-strict-mode.md **Status**: Keep **Reasoning**: All references verified. `tsconfig.app.json` confirms ES2020 target with `lib: ["ES2020"]`. The `tsgo` installation note (Go binary, not npm package) and `response.json()` returning `unknown` are valuable gotchas. Node.js >= 24 requirement is noted. --- #### rules/openai-reasoning-models.md **Status**: Needs Update **Reasoning**: The core concept is still valid - orphaned reasoning parts are still filtered in `src/ipc/utils/ai_messages_utils.ts`. However, the specific function name referenced is outdated. **Issues Found**: - References `filterOrphanedReasoningParts()` as a named function, but this logic was refactored into the `cleanMessage()` function (inline filtering within that function). The named export no longer exists. **Suggested Changes**: - Update the reference from `filterOrphanedReasoningParts()` to describe the filtering logic within `cleanMessage()` in `src/ipc/utils/ai_messages_utils.ts` --- #### rules/adding-settings.md **Status**: Keep **Reasoning**: All file paths verified: `UserSettingsSchema` in `src/lib/schemas.ts`, `DEFAULT_SETTINGS` in `src/main/settings.ts`, `SETTING_IDS` in `src/lib/settingsSearchIndex.ts`, `AutoApproveSwitch.tsx` as template. Recent commit d6ab829 (add max tool call steps setting) confirms this pattern is actively used. --- #### rules/chat-message-indicators.md **Status**: Keep **Reasoning**: Short (12 lines), low token cost. `dyad-status` tag implementation confirmed in

upstream(agents): 移植 5 个冲突 commit (P0,P1) — v2026.3.7→v2026.3.8

## 任务 将以下 5 个上游 commit 的修改语义化应用到本 fork。这些 commit 无法直接 cherry-pick（存在冲突），需要理解修改意图后手动应用等效变更。 ### 上游版本范围 - **来源**: openclaw/openclaw v2026.3.7 → v2026.3.8 - **模块**: `agents` - **优先级**: P0,P1 ### 需要移植的 commit #### Commit 1: `e8775cda932f` (P1) **描述**: fix(agents): re-expose configured tools under restrictive profiles **涉及文件**: `src/agents/pi-tools.policy.test.ts,src/agents/pi-tools.policy.ts,src/plugins/config-state.test.ts,src/plugins/config-state.ts` <details> <summary>查看上游 diff</summary> ```diff diff --git a/src/agents/pi-tools.policy.test.ts b/src/agents/pi-tools.policy.test.ts index 4b7a16b4d..0cdc572c4 100644 --- a/src/agents/pi-tools.policy.test.ts +++ b/src/agents/pi-tools.policy.test.ts @@ -3,6 +3,7 @@ import type { OpenClawConfig } from "../config/config.js"; import { filterToolsByPolicy, isToolAllowedByPolicyName, + resolveEffectiveToolPolicy, resolveSubagentToolPolicy, } from "./pi-tools.policy.js"; import { createStubTool } from "./test-helpers/pi-tool-stubs.js"; @@ -176,3 +177,59 @@ describe("resolveSubagentToolPolicy depth awareness", () => { expect(isToolAllowedByPolicyName("sessions_spawn", policy)).toBe(false); }); }); + +describe("resolveEffectiveToolPolicy", () => { + it("implicitly re-exposes exec and process when tools.exec is configured", () => { + const cfg = { + tools: { + profile: "messaging", + exec: { host: "sandbox" }, + }, + } as OpenClawConfig; + const result = resolveEffectiveToolPolicy({ config: cfg }); + expect(result.profileAlsoAllow).toEqual(["exec", "process"]); + }); + + it("implicitly re-exposes read, write, and edit when tools.fs is configured", () => { + const cfg = { + tools: { + profile: "messaging", + fs: { workspaceOnly: false }, + }, + } as OpenClawConfig; + const result = resolveEffectiveToolPolicy({ config: cfg }); + expect(result.profileAlsoAllow).toEqual(["read", "write", "edit"]); + }); + + it("merges explicit alsoAllow with implicit tool-section exposure", () => { + const cfg = { + tools: { + profile: "messaging", + alsoAllow: ["web_search"], + exec: { host: "sandbox" }, + }, + } as OpenClawConfig; + const result = resolveEffectiveToolPolicy({ config: cfg }); + expect(result.profileAlsoAllow).toEqual(["web_search", "exec", "process"]); + }); + + it("uses agent tool sections when resolving implicit exposure", () => { + const cfg = { + tools: { + profile: "messaging", + }, + agents: { + list: [ + { + id: "coder", + tools: { + fs: { workspaceOnly: true }, + }, + }, + ], + }, + } as OpenClawConfig; + const result = resolveEffectiveToolPolicy({ config: cfg, agentId: "coder" }); + expect(result.profileAlsoAllow).toEqual(["read", "write", "edit"]); + }); +}); diff --git a/src/agents/pi-tools.policy.ts b/src/agents/pi-tools.policy.ts index db9a36755..61d037dd9 100644 --- a/src/agents/pi-tools.policy.ts +++ b/src/agents/pi-tools.policy.ts @@ -2,6 +2,7 @@ import { getChannelDock } from "../channels/dock.js"; import { DEFAULT_SUBAGENT_MAX_SPAWN_DEPTH } from "../config/agent-limits.js"; import type { OpenClawConfig } from "../config/config.js"; import { resolveChannelGroupToolsPolicy } from "../config/group-policy.js"; +import type { AgentToolsConfig } from "../config/types.tools.js"; import { normalizeAgentId } from "../routing/session-key.js"; import { resolveThreadParentSessionKey } from "../sessions/session-key-utils.js"; import { normalizeMessageChannel } from "../utils/message-channel.js"; @@ -196,6 +197,37 @@ function resolveProviderToolPolicy(params: { return undefined; } +function resolveExplicitProfileAlsoAllow(tools?: OpenClawConfig["tools"]): string[] | undefined { + return Array.isArray(tools?.alsoAllow) ? tools.alsoAllow : undefined; +} + +function hasExplicitToolSection(section: unknown): boolean { + return section !== undefined && section !== null; +} + +function resolveImplicitProfileAlsoAllow(params: { + globalTools?: OpenClawConfig["tools"]; + agentTools?: AgentToolsConfig; +}): string[] | undefined { + const implicit = new Set<string>(); + if ( + hasExplicitToolSection(params.agentTools?.exec) || + hasExplicitToolSection(params.globalTools?.exec) + ) { + implicit.add("exec"); + implicit.add("process"); + } + if ( + hasExplicitToolSection(params.agentTools?.fs) || + hasExplicitToolSection(params.globalTools?.fs) + ) { + implicit.add("read"); + implicit.add("write"); + implicit.add("edit"); + } + return implicit.size > 0 ? Array.from(implicit) : undefined; +} + export function resolveEffectiveToolPolicy(params: { config?: OpenClawConfig; sessionKey?: string; @@ -226,6 +258,15 @@ export function resolveEffectiveToolPolicy(params: { modelProvider: params.modelProvider, modelId: params.modelId, }); + const explicitProfileAl

X Users Find Their Real Names Are Being Googled in Israel After Using X Verification Software “Au10tix”

X Users Find Their Real Names Are Being Googled in Israel After Using X Verification Software “Au10tix” Alan Macleod On January 30, the Department of Justice released its latest tranche of 3.5 million documents relating to Jeffrey Epstein. Years of emails, texts, and images were suddenly in the public domain. Epstein, a serial rapist, masterminded a global human trafficking and sexual abuse network, and could count princes, professors, and politicians among his closest friends and accomplices. MintPress News has been at the forefront of covering the Epstein saga, revealing his extremely close links to American and Israeli intelligence groups – a discovery that perhaps sheds light on why it took so long for the world’s most notorious pedophile to face accountability for his crimes. Many of the DOJ files have been heavily redacted in order to protect Epstein’s powerful clients. Still, they have exposed a massive elite nexus revolving around the New York billionaire, implicating presidents, diplomats, and plutocrats in his crimes, and imply that Epstein was significantly more powerful than first thought, shaping modern politics in ways never previously understood. With shocking new details emerging on a near-hourly basis, here are ten Epstein- related stories that have flown relatively under the radar. The Israeli Government Installed Surveillance Cameras at Epstein’s New York Apartment The Israeli government installed and maintained a hi-tech surveillance system at Epstein’s Manhattan apartment complex, including a network of alarms and cameras, emails show. Starting in 2016, the director of protective service at the Israeli mission to the United Nations controlled guests’ access to the Manhattan residence, and even performed background checks on prospective cleaners and other Epstein employees. Former Israeli prime minister Ehud Barak admitted visiting the apartment up to 100 times, and stayed there for long periods of time. While Barak’s security may have been a concern, Epstein is known to have housed underage girls at the apartment, and many of his worst sexual crimes and most sordid parties were held there, raising questions as to what sort of images and data the Israeli government had access to. Epstein Plotted War With Iran Ehud Barak became one of Epstein’s closest associates, staying for extended periods of time at the billionaire’s residences. The pair would email, text, call, and meet constantly. A search for “Ehud Barak” elicits more than 3500 results in the latest file dump alone. The pair would talk politics, and shared a vision of the United States attacking Iran. In 2013, with negotiations between the International Atomic Energy Agency and Iran stalling, Epstein emailed Barak stating, in typically poor spelling and grammar: “hopefully somone suggests getting authorization now for Iran. the congress woudl do it.” Epstein would get his wish in 2025, when his close associate Donald Trump began bombing the country. Noam Chomsky Considered Epstein His “Best Friend” Epstein arranged a meeting between Barak and renowned leftist academic (and vehement critic of the U.S. and Israel) Noam Chomsky. An unlikely friendship between the notorious pedophile and star professor blossomed, with the pair regularly meeting up at each other’s houses for dinner. Chomsky flew on Epstein’s “Lolita Express” jet to attend a dinner with Woody Allen in New York. He also expressed his desire to visit Little St. James Island, Epstein’s notorious Caribbean hideaway, and the center of his trafficking operation. Chomsky considered Epstein his “best friend” according to an email sent by his wife, Valeria. The usually curt and matter-of-fact academic signed off his emails to Epstein with unexpectedly flowery language, such as “Like real friendship, deep and sincere and everlasting from both of us, Noam and Valeria.” Chomsky strongly supported Epstein until his dying day in a Manhattan prison cell, taking it upon himself to act as his unofficial crisis manager, describing his accusers as “publicity seekers or cranks of all sorts,” and denouncing the media as a “culture of gossip-mongers” destroying his stellar character. “Ive watched the horrible way you are being treated in the press and public,” he wrote, advising Epstein on tactics to fight the supposed smears against him. For a full rundown of the Chomsky-Epstein relationship, see the MintPress News investigation: “The Chomsky-Epstein Files: Unravelling a Web of Connections Between a Star Leftist Academic and a Notorious Pedophile.” Steve Bannon Developed a Plan to Help Epstein “Crush the Pedo Narrative” A second public figure running defense for Epstein was Steve Bannon. In public, the far-right strategist claimed that he was working on a documentary exposing Epstein. In private messaging, however, Bannon, like Chomsky, was advising Epstein on how best to repair his image. Just weeks before Epstein’s arrest and subsequent death, Bannon was messaging him, devising a complex media strategy

Extra! Extra! 3/8

[](https://substackcdn.com/image/fetch/$s_!QSz6!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Ff77f594e-6606-41d3-a0d0-47208fab47b5_1488x1348.jpeg) Spotted in Jay Kuo’s “[Just for Skeets and Giggles](https://statuskuo.substack.com/p/just-for-skeets-and-giggles-3726).” Hi, all, and happy Sunday. It’s been another bear of a week, so it’s more important than ever to stop and take in the remarkable amount of good news we ALSO had. Remember, it’s not only OK to take a break from doomscrolling to celebrate these wins—it’s necessary. Every army needs the occasional morale boost; ours is no exception. So read this list, please, and share it with others who feel like “nothing good is happening.” In fact quite a bit is, but folks won’t know about it if we don’t spread the word. Your hard work matters. It’s the reason these good news lists come out every week. So enjoy this post, celebrate it…and know that you’re the reason we’ll have another one next week. ## Celebrate This! 🎉 Dr. Vinay Prasad, the FDA’s polarizing vaccine chief, [is leaving the agency](https://apnews.com/article/vinay-prasad-fda-vaccines-laura-loomer-83030ad6eb7651095e3c40444dd69f12). Cook County prosecutors [dismissed 21 cases](https://article.wn.com/view-lemonde/2026/03/04/Cook_County_prosecutors_dismiss_21_cases_against_ICE_protest/#/related_news) that were filed against protesters at the ICE processing center in Broadview, IL, including charges for 15 moms who hopped concrete barricades in a highly-publicized act of civil disobedience. A federal judge [ruled](https://www.nytimes.com/2026/03/07/us/politics/judge-kari-lake-voa-layoffs.html) that the appointment of Kari Lake, the head of Voice of America’s oversight agency, was invalid, voiding mass layoffs that she had carried out at the federally funded news group last year. Support for abolishing ICE has hit a new high in [this week's Economist / YouGov poll](https://d3nkl3psvxxpe9.cloudfront.net/documents/econTabReport_ubu5DXD.pdf#page=37). Half (50%) of Americans now somewhat or strongly support abolishing ICE. Only 39% oppose abolishing the agency. In Los Angeles, a mobile clinic [is now bringing mammograms](https://www.yahoo.com/news/articles/mobile-clinic-brings-mammograms-women-110000308.html?ck_subscriber_id=2496857656&guccounter=1) to women on Skid Row. Oregon state Democratic lawmakers [approved a measure that would prevent federal immigration officers from wearing masks](https://www.opb.org/article/2026/03/05/oregon-lawmakers-approve-measure-prohibit-masks-ice-agents/). A federal court [ordered the Trump administration](https://ctmirror.org/2025/12/12/fema-funding-bric-judge-orders-restored/) to restore billions in funding for the Building Resilient Infrastructure and Communities (BRIC) program, which had been canceled. This ruling came after a lawsuit from 22 Democratic Attorneys General. EVs are officially [making the air cleaner.​](https://grist.org/solutions/evs-are-already-making-your-air-cleaner/?sh_kit=7a2950363f4b90b1881ae76c68d24551846eea9063b67a6a14e9fa39bc419e40) A Minnesota prosecutor [said her office in Hennepin County is investigating the "potentially ​unlawful behavior" of federal agents](https://www.reuters.com/legal/government/top-border-patrol-official-other-federal-agents-being-investigated-by-2026-03-02/?sh_kit=7a2950363f4b90b1881ae76c68d24551846eea9063b67a6a14e9fa39bc419e40), including Gregory Bovino, during the ICE surge earlier this year. New York City Mayor Zohran Mamdani and New York Governor Kathy Hochul [announced the first neighborhoods in the city that will get free childcare for 2-year-olds](https://9905ebc8.click.kit-mail3.com/gku2xo2do4b5hl9l76wtrh8elgvx5cnm6o68d7p782zg636ozlk9ew0xwvwmgeqzz3r780kdnqvevm69r07573kpzx56nkz4m8p25ne22kq3dk5rnexddzln745h9p9e5/58hvh7hgqdnzmga7/aHR0cHM6Ly9hYmM3bnkuY29tL3Bvc3QvbWFtZGFuaS1ob2NodWwtdW52ZWlsLWZpcnN0LW55Yy1uZWlnaGJvcmhvb2RzLWdldC1mcmVlLWNoaWxkLWNhcmUtMi15ZWFyLW9sZHMtZmFsbC8xODY3MTQ4NS8=). [More people in rural majority-Latino TX counties turned out to vote in the Democratic primary](https://bsky.app/profile/tonolatino.bsky.social/post/3mgbtqsico22y) than the number of people who voted for Harris in 2024. GOP Rep. Tony Gonzales [dropped his re-election bid](https://ground.news/article/gop-rep-tony-gonzales-drops-re-election-bid-amid-ethics-probe-into-his-affair-with-a-staffer_dc69c3) amid an ethics probe into his affair with a staffer. Rep. Burgess Owens (R-Utah) [will not seek reelection this November](https://www.axios.com/local/salt-lake-city/2026/03/04/burgess-owens-wont-seek-reelection?emci=9406156c-bd19-f111-a69a-000d3a1

Recall vs. Wisdom: What Over-Personalization Reveals About the Future of Relational AI

[Original Reddit post](https://www.reddit.com/r/ArtificialInteligence/comments/1ro4k19/recall_vs_wisdom_what_overpersonalization_reveals/) The over-personalization problem isn’t really about memory. It’s about relationship. When an AI assistant drags your hiking preferences into a weather query, the failure isn’t technical recall gone haywire. It’s a system that has no idea what it means to actually be in a conversation with someone. That distinction matters more than it might seem, because the entire industry just bet big on the opposite assumption. Google recently rolled out automatic memory for Gemini. The feature is on by default. Without any prompting from the user, Gemini now recalls “key details and preferences” from past conversations and injects them into future responses. Google frames this as “Personal Intelligence,” a system that connects the dots across Gmail, Photos, Search, and YouTube to make the assistant “uniquely helpful for you.” And it’s not just Gemini. This is part of a broader push to make memory the centerpiece of the AI assistant experience. The pitch is simple: the more an AI knows about you, the better it serves you. But OP-Bench, the first systematic benchmark for over-personalization, tells a different story. It turns out that the more aggressively a system uses what it remembers, the worse the interaction gets. Not occasionally. Universally. Every memory-augmented system they tested showed severe over-personalization. And the more sophisticated the memory architecture, the harder it failed. We’ve been so focused on the capacity to remember that we’ve neglected the wisdom of when to use what we remember. That’s not an engineering oversight. It’s a relational one. Memory Without Attunement Is Just Surveillance Here’s the thing. A system that remembers everything about you and surfaces it indiscriminately isn’t being helpful. It’s performing ambient surveillance dressed up as personalization. People describe over-personalizing systems as “creepy” and “overly familiar,” and those aren’t technical complaints. They’re relational ones. The system has violated something unspoken about when personal knowledge should enter a conversation. Google’s approach makes this tension vivid. Gemini doesn’t just remember what you explicitly told it to remember. It silently mines your past conversations for details and preferences, then weaves them into future responses without asking whether that’s what you wanted. The feature shipped turned on by default. You have to go dig through Settings, find “Personal context,” and manually toggle it off. If you’re a Google AI Pro or Ultra subscriber, the “Personal Intelligence” layer goes further, pulling context from your email, your photos, your search history. The integration is seamless, which is exactly what makes it concerning. This maps onto one of the foundational problems in relational AI: the difference between knowing about someone and being attuned to them. Knowing about someone is a database operation. You store facts, retrieve them, insert them into responses. Attunement is qualitatively different. It requires reading the current moment, understanding what the person actually needs right now, and making a judgment call about which pieces of shared history belong in this exchange and which ones don’t. OP-Bench makes this distinction measurable for the first time. Their three failure modes map cleanly onto relational breakdowns. Irrelevance is a failure of contextual reading: the system can’t tell the difference between “semantically similar” and “conversationally appropriate.” Sycophancy is a failure of honesty: the system weaponizes personal knowledge to tell you what you want to hear instead of what’s true. Repetition is a failure of presence: the system is stuck rehashing old interactions instead of engaging with this one. All three are failures of attunement, not memory. The Attention Hijack The technical finding about “memory hijacking” deserves a closer look. When researchers examined attention patterns, they found that memory-augmented models attend to retrieved memory tokens at roughly twice the rate they attend to the actual user query. Let that sink in. The model is paying more attention to what it already knows about you than to what you’re saying right now. In any healthy relationship, the balance between history and presence matters. You bring what you know about the other person into the conversation, but you don’t let it drown out your ability to listen. Over-personalizing systems have lost that balance entirely. They’re so saturated with stored context that they can’t hear the present moment. And this isn’t just a chatbot problem. As we build multi-agent systems where AI agents maintain persistent memory about users, tasks, and each other, the attention hijacking problem scales in ways that should worry anyone thinking about agent coordination. An agent that over-attends to stored context about another agent’s past behavior wil

DeBriefed 6 March 2026: Iran energy crisis

W*elcome to Carbon Brief’s DeBriefed.* *An essential guide to the week’s key developments relating to climate change.* # **This week** ### **Energy crisis** **ENERGY SPIKE:** US-Israeli attacks on Iran and subsequent counterattacks across the Middle East have sent energy prices “soaring”, according to [Reuters](https://www.reuters.com/business/energy/global-energy-costs-soar-iran-crisis-disrupts-shipping-oil-gas-production-2026-03-03/). The newswire reported that the region “accounts for just under a third of global oil production and almost a fifth of gas”. The [Guardian](https://www.theguardian.com/world/2026/mar/02/iran-strait-of-hormuz-oil-gas-visualized?) noted that shipping traffic through the strait of Hormuz, which normally ferries 20% of the world’s oil, “all but ground to a halt”. The [Financial Times](https://www.ft.com/content/dac7a77d-e0f4-4f52-a3d4-55b145e67347) reported that attacks by Iran on Middle East energy facilities – notably in Qatar – triggered the “biggest rise in gas prices since Russia’s full-scale invasion of Ukraine”. **‘RISK’ AND ‘BENEFITS’:** [Bloomberg](https://www.bloomberg.com/news/articles/2026-03-03/global-diesel-prices-surge-higher-as-iran-war-disrupts-supplies) reported on increases in diesel prices in Europe and the US, speculating that rising fuel costs could be “a risk for president Donald Trump”. US gas producers are “poised to benefit from the big disruption in global supply”, according to [CNBC](https://www.cnbc.com/2026/03/03/us-natural-gas-lng-qatar-iran-war.html). Indian government sources told the [Economic Times](https://pdpwbj.clicks.mlsend.com/tl/c/eyJ2Ijoie1wiYVwiOjI0OTYxNyxcImxcIjoxODEwMDA5MzYwMDg3MTM4MjQsXCJyXCI6MTgxMDAwOTQ5MjYxNjY1ODA5fSIsInMiOiI4N2E5OWQ3ZTZiNDg0OTRlIn0) that Russia is prepared to “fulfil India’s energy demands”. [China Daily](https://www.chinadaily.com.cn/a/202603/03/WS69a64540a310d6866eb3b4a2.html) quoted experts who said “China’s energy security remains fundamentally unshaken”, thanks to “emergency stockpiles and a wide array of import channels”. **‘ESSENTIAL’ RENEWABLES:** Energy analysts said governments should cut their fossil-fuel reliance by investing in renewables, “rather than just seeking non-Gulf oil and gas suppliers”, reported [Climate Home News](https://www.climatechangenews.com/2026/03/04/gulf-oil-and-gas-crisis-sparks-calls-for-renewable-invesment). This message was echoed by UK business secretary Peter Kyle, who said “doubling down on renewables” was “essential” amid “regional instability”, according to the [Daily Telegraph](https://www.telegraph.co.uk/business/2026/03/03/net-zero-answer-middle-east-energy-crisis/). ### **China’s climate plan** **PEAK COAL?:** China has set out its next “five-year plan” at the annual “[two sessions](https://pdpwbj.clicks.mlsend.com/td/cl/eyJ2Ijoie1wiYVwiOjI0OTYxNyxcImxcIjoxODEwOTE4NDc3Nzc1NTIyNDAsXCJyXCI6MTgxMDkxODYxODA4NTQ2OTgyfSIsInMiOiIzZDZmMjQyY2JiMmIzNTM3In0)” meeting of the National People’s Congress, including its climate strategy out to 2030, according to the Hong Kong-based [South China Morning Post](https://www.scmp.com/economy/china-economy/article/3345525/china-step-tech-energy-and-decarbonisation-efforts-next-5-year-plan). The plan called for China to cut its carbon emissions per unit of gross domestic product (GDP) by 17% from 2026 to 2030, which “may allow for continued increase in emissions given the rate of GDP growth”, reported [Reuters](https://www.reuters.com/sustainability/climate-energy/china-plans-cut-carbon-dioxide-emissions-per-unit-gdp-by-around-38-2026-2026-03-05/). The newswire added that the plan also had targets to reach peak coal ​in the next five years and replace 30m tonnes per year of coal with renewables. **ACTIVE YET PRUDENT:** [Bloomberg](https://www.bloomberg.com/news/articles/2026-03-05/china-aims-to-cut-carbon-emissions-per-unit-of-gdp-17-by-2030) described the new plan as “cautious”, stating that it “frustrat[es] hopes for tighter policy that would drive the nation to peak carbon emissions well before president Xi Jinping’s 2030 deadline”. Carbon Brief has just published an in-depth [analysis](https://www.carbonbrief.org/qa-what-does-chinas-15th-five-year-plan-mean-for-climate-change/) of the plan. [China Daily](https://www.chinadaily.com.cn/a/202603/05/WS69a91c1ba310d6866eb3be81.html) reported that the strategy “highlights measures to promote the climate targets of peaking carbon dioxide emissions before 2030”, which China said it would work towards “actively yet prudently”. # **Around the world** **EU RULES:** The European Commission has proposed new “made in Europe” rules to support domestic low-carbon industries, “against fierce competition from China”, reported [Agence France-Presse](https://www.france24.com/en/live-news/20260304-eu-to-unveil-made-in-europe-rules-despite-pushback). [Carbon Brief](https://www.carbonbrief.org/qa-what-the-eus-new-industry-and-made-in-europe-rules-mean-for-climate-action/) examined what it means for c

Sen. Sheldon Whitehouse (D-RI) lays out the connections between Trump, Russia, and Epstein (transcript included)

**NOTE:** This transcript now appears in [the Senate section of the official *Congessional Record* of March 5, 2026, pages 18 - 23,](https://www.congress.gov/119/crec/2026/03/05/172/42/CREC-2026-03-05-senate.pdf) with Sen. Whitehouse's own list of sources appended. ----- The following is the YouTube transcript which I cleaned up, checked for errors, lightly edited for readability, verified spelling of proper names via Wikipedia, and added links to any quotes that I checked myself. (EDITED to add links to individuals mentioned, correct placement of quotes, and insert links to original articles where I could find them online) I found myself doing it anyway just for me, to keep track of who's who, and then I realized I might as well do it for you as well. This is an unparalleled speech: while the substance of it might be available elsewhere and I've just missed it, Sen. Whitehouse has answered a lot of questions in my mind about not just the links between Trump, Russia, and Epstein -- and William Barr as one of many links -- but also about the recording equipment and blackmail angle that is present in so many survivor accounts and so noticeably absent everywhere else. It's truly worth listening to, but if you can't sit still that long, here's the transcript. ----- Thank you, Madam President. It was the spring of 2019. Public and media interest in special counsel [Robert Mueller's report into Russia's election interference operation](https://en.wikipedia.org/wiki/Mueller_special_counsel_investigation) reached a fever pitch. There had been a steady drip, drip, drip of reporting on the Trump team's cozy and peculiar relationship with Russia. Since his surprise election victory in 2016, ahead of the Mueller report's release, Trump's Attorney General, Bill Barr, [issued a letter to Congress purporting to summarize the report's findings.](https://en.wikipedia.org/wiki/Barr_letter) The letter declared that Russia and the Trump campaign did not collude to steal the election. The press, ravenous for any news of the long-anticipated Mueller report's conclusion, largely accepted [Attorney General Barr's](https://en.wikipedia.org/wiki/William_Barr) narrow, carefully worded conclusion and, not yet having access to the full report, blasted the attorney general's summary around the world. Trump himself declared, all caps, NO COLLUSION. He said he had been cleared of the Russia "hoax," a term he reserves only to describe things that are true, like climate change. Frustrated, Mueller wrote to Barr that the attorney general's letter did not fully capture the context, nature, and substance of the investigation. But by the time [the dense, voluminous Mueller report](https://en.wikipedia.org/wiki/Mueller_report) was issued the month after Barr's letter, its message had been obscured. The Mueller report actually concluded that the Trump campaign knew of and welcomed Russian interference and expected to benefit from it. That conclusion was later echoed and reinforced by [an investigation led by then-chairman Marco Rubio's Senate Intelligence Committee,](https://en.wikipedia.org/wiki/Mueller_report#Senate_Intelligence_Committee) a bipartisan report. But Barr's scheme had largely worked. Many in the media and in the Democratic Party seemed to internalize that the Russia speculation had perhaps gotten out of hand, and that perhaps we had been wrong to believe there was a troubling connection between Trump and Russia after all. But were we? Let's take a look at a sampling of what Trump has done for Russia just lately, and usually at the expense of American interests. There are many, but here's a top 10. **One,** after Trump and Vice President Vance theatrically chastised the heroic Ukrainian President Zelenskyy in front of TV cameras in the Oval Office last year, Trump paused our weapons shipments to Ukraine. **Two,** in July, during the worst Russian bombing campaign of the war until that point, Trump paused an already funded weapons shipment for Ukraine, including the Patriot interceptors that protect civilians from Putin's savage attacks. **Three,** that same month, Trump's Treasury Department stopped imposing new sanctions and closing sanctions loopholes, effectively allowing dummy corporations to send funds, chips, and military equipment to Russia. **Four,** leaked phone calls show that White House envoy [Steve Witkoff](https://en.wikipedia.org/wiki/Steve_Witkoff) and Putin envoy [Kirill Dmitriev](https://en.wikipedia.org/wiki/Kirill_Dmitriev) have worked together closely behind the scenes on a peace deal favorable to Russia. **Five,** last summer, Trump rolled out the presidential red carpet for the Russian dictator on American soil. with a summit in Alaska that yielded unsurprisingly no gains toward ending the war in Ukraine. **Six,** Trump's vice president traveled to the Munich Security Conference last year to parrot Russia's anti-western talking points pushed by right-wing groups that Puti