Eli Clifton speaks to Mondoweiss about the upcoming U.S. midterm elections, the shifting landscape of AIPAC, and his new book, Israel’s Lobby: America in the Grip of a Foreign Power.

Russian forces have been unable to make gains near Hulyaipole and are shifting their focus to other sectors, but Ukrainian troops are also facing challenges in the area, according to Major Dmytro Filatov, commander of the 1st Separate Assault Brigade. Filatov said Ukrainian forces lack the weaponry needed to consistently disrupt Russian logistics deep behind enemy lines, even though similar operations brought significant success over the summer. Filatov reported that Russian troops are not seeing major advances in Zaporizhzhia region, with Ukrainian forces having stabilized the situation in the area. Speaking on Hromadske Radio, he said there have been occasional instances where Russian forces manage to push forward, but these are rare, and the Defense Forces usually counter them quickly, sometimes even improving their own positions. “These are not on a large scale, and it’s not deep penetration,” Filatov said. “The line of contact resembles a small wave, rolling back and forth. These are inevitable dynamics in combat, but the enemy hasn’t been able to achieve any significant success, which is why they’re refocusing on other sectors.” However, Ukrainian forces are also encountering problems on this front. One issue, he explained, is their limited ability to strike Russian logistics targets at operational depth with consistency. Ukrainian units have found ways to conduct such operations, but cannot sustain the same tempo of strikes. Over the summer, they managed to effectively block vehicle movements from Tokmak and Melitopol, which impacted Russian military supplies. “The occupation authorities even banned people from leaving the cities. That allowed us to improve the situation at the front line, because the enemy stopped resupplying its units with explosives and hasn’t been giving infantry the means it needs,” Filatov recalled. But due to a lack of resources, the scale of these operations has now decreased. According to Filatov, today Ukrainian troops are carrying out only about 5% of the number of strikes they managed during the summer. In late June 2026, DeepState analysts said Russian troops had fully occupied Hulyaipole in Zaporizhzhia region, which had served as a key defensive stronghold for Ukraine in the south. Although the General Staff has not officially confirmed this, the city appears as controlled by Russian forces on DeepState’s interactive map. Hulyaipole is considered important to Russia because it is a major logistics hub, as is the nearby town of Zaliznychne, which is partially held by Russian troops. The occupation of the city also opened up possible routes for Russian advances toward Zaporizhzhia, Orikhiv, and even into the rear of Ukrainian positions in Donetsk region, particularly near Pokrovsk and Velyka Novosilka.

We are exploring how to make peace with shifting priorities and move forward without guilt when lifelong friendships begin to fade.

Cybersecurity teams are shifting toward continuous control across identity, cloud, endpoints, telemetry, and human risk.

The 30-Second Rundown * **Enterprise buyers are shortening AI SaaS contracts while foundational AI labs commit over $500 billion to physical compute infrastructure.** — Enterprise IT spend is shifting from locked-in software subscriptions directly toward custom hardware and dynamic model orchestration layers. * **Developers can now install natural-language prompt cache keepers and session handoff tools inside AI coding environments to cut token costs.** — Prevents context degradation and saves thousands in developer API usage through automated prompt caching management. * **Microsoft and OpenAI are pivoting toward ambient computer-use agents and deep software integration over pure model intelligence.** — Distribution and enterprise contextual data access are replacing foundational model benchmark scores as the primary enterprise moat. * **Engineering teams are implementing automated model routing layers to dynamically balance workloads between high-cost reasoning models and fast open-source options.** — Dramatically reduces inference costs by reserving expensive top-tier models strictly for ambiguous, complex logical tasks. ## Guru Chatter ### Erosion of Multi-Year AI SaaS Contracts and $500 Billion Compute Hyper-CapEx **TL;DR:** Big corporate buyers are refusing long-term software contracts because AI models change too fast. At the same time, top AI companies are spending hundreds of billions building massive data centers to secure hardware. Enterprise software buyers are actively resisting long-term Annual Recurring Revenue (ARR) commitments with model providers due to rapid capability shifts and open-source competition. Conversely, frontier AI labs are securing unprecedented physical infrastructure commitments ($500B+) ahead of potential public offerings, treating raw hardware fabric and power access as the primary strategic moat. **Market impact:** Destabilizes traditional ARR valuation metrics for top-layer SaaS wrappers while driving massive capital allocation into custom silicon, server fabricators, and energy infrastructure. Portfolios should reweight toward compute infrastructure, physical energy backbones, and model-agnostic middleware. **Sources:** AI News & Strategy Daily | Nate B Jones * * * ### Enterprise Agent Distribution and Ambient Computer-Use Execution **TL;DR:** AI tools are shifting from simple chat boxes into background agents that run continuously inside productivity apps or control web browsers directly. Distribution via established productivity suites (e.g., Microsoft 365 Copilot/Autopilot) and ambient computer-use layers (e.g., ChatGPT Dots) is outpacing raw foundational LLM performance. Continuous, background-scheduled execution engines and organizational context access are becoming the core determinants of commercial software adoption. **Market impact:** Reduces enterprise reliance on standalone vertical SaaS tools and custom web scraping pipelines. Favors platforms with native computer-use capabilities and direct enterprise context access, shifting high-concurrency compute demand toward continuous 24/7 background agent execution. **Sources:** AI News & Strategy Daily | Nate B Jones · The AI Advantage * * * ### Context Economics and Active Prompt Cache Lifecycle Orchestration **TL;DR:** AI coding tools are giving developers real-time dashboards to track context window limits, cache expiration timers, and money spent per session. Developer tooling ecosystems are evolving from passive context management to active lifecycle orchestration. Interfaces now explicitly track prompt caching Time-To-Live (TTL) limits (e.g., Anthropic's 60-minute window) and surface context saturation metrics directly, enabling programmatic context clearing, automated handoffs, and state compression before cache degradation occurs. **Market impact:** Dramatically lowers total cost of ownership for enterprise LLM API usage. Drives capital toward intelligent agentic middleware, client-side state managers, and cache-aware orchestration layers that eliminate redundant token ingestion. **Sources:** Nate Herk | AI Automation · Nate Herk | AI Automation * * * ### Dynamic Multi-Tier Model Routing and Task-Specific Execution Asymmetry **TL;DR:** Smart apps now use cheap, fast AI models for simple background tasks and automatically switch to expensive reasoning models only when complex decision-making is required. Architectures are increasingly decoupling basic data processing from complex strategic reasoning. Multi-tier routers analyze query parameters, context size, and required logical depth to route micro-tasks dynamically between local open-source models, fast mid-tier models (Claude Sonnet), and heavy flagships (Claude Opus), minimizing token costs without sacrificing output quality. **Market impact:** Reduces margin compression for enterprise AI deployments by optimizing inference unit economics. Shifts software infrastructure spend toward orchestration gateways, semantic intent classifiers, and smaller, domain-specialized execution models. **Sources:** AI News & Strategy Daily | Nate B Jones · Nate Herk | AI Automation * * * ### Intent-Driven Real-Time Prototyping and Generalist Product Engineering **TL;DR:** Software development is shifting from writing precise technical blueprints to describing high-level goals, allowing individual builders to manage end-to-end stack creation in real time. Software engineering is transitioning from specification-heavy development toward Intent-Driven Prototyping. Autonomous AI agents allow single generalist engineers to manage frontend UI, backend APIs, database authorization, and cloud hosting natively, compressing traditional multi-specialist software engineering cycles into rapid, real-time feedback loops. **Market impact:** Significantly reduces developer headcount requirements for early-stage software development. Tech capital shifts away from monolithic task management software toward agentic execution environments, rapid preview hosting providers, and automated code-guardrail middleware. **Sources:** Nate Herk | AI Automation * * * ### Transition from Token Bottlenecks to Human Executive Function Constraints **TL;DR:** As AI token limits and raw speeds grow exponentially, the real limiting factor is no longer computing power, but human time to check and review AI output. With high-throughput tiers and expanding context limits, model availability is no longer the primary operational constraint. System bottlenecks have migrated to human executive function—specifically the cognitive bandwidth required to review, verify, and direct synthetic output without creating unvetted digital noise. **Market impact:** Reallocates software value creation from pure text/code generation toward automated output verification, synthesized curation frameworks, and high-trust human-in-the-loop review orchestration software. **Sources:** David Shapiro * * * ### Graph-Based Relationship Networks over Cold Marketplaces **TL;DR:** Standard job applications on public boards are failing due to automated applicant spam, forcing organizations to rely on automated graph networks and warm internal referrals. Public talent marketplaces suffer from a high noise floor caused by automated application submission tools. High-trust network graphs (alumni connections, direct relationships, secondary ties) processed via Graph Neural Networks (GNNs) are superseding generic applicant tracking systems. **Market impact:** Accelerates software adoption of graph databases and relationship resolution models in HR technology, driving enterprise spend away from generic open job boards toward private graph-based talent verification networks. **Sources:** Joshua Fluke ## Master Workflows Today's Top Pick ### Claude Code Context Cache Keeper and Automated Session Handoff Intermediate~20 min **Why it's worth it:** Eliminates repetitive API context costs, prevents context window rot, and preserves continuous developer session state automatically. Integrates a custom status-bar mod into Claude Code that tracks token saturation, prompt cache expiration timers (60-minute TTL), and API costs in real time. Upon cache expiration warnings, it triggers an automated context summarization skill and resets the workspace context window cleanly. Claude Code CLIClaude Code Desktop AppAnthropic Claude APIPrompt Caching Infrastructure 1. Initialize the plugin configuration within the Claude Code terminal or desktop interface. /plugin 2. Prompt Claude Code to generate a custom status-bar telemetry mod named Cache Keeper that tracks session usage, token budget, and prompt cache expiration. Create a status-bar mod named 'Cache Keeper' that monitors active session token count, context percentage, estimated API billable cost, and a 60-minute prompt cache TTL countdown timer. Send a desktop notification when the cache timer drops below 5 minutes. 3. Add dynamic handoff triggers to the mod UI to generate a structured context summary prior to resetting the session. Add an automated handoff button that runs a workspace summary skill, generating a concise Markdown brief of current task status, open bugs, and next implementation steps. 4. Execute the clear command in the terminal to clear context memory and paste the generated handoff brief into the fresh session. claude --clear **Links:** https://docs.anthropic.com/en/docs/build-with-claude/prompt-caching **Sources:** Nate Herk | AI Automation · Nate Herk | AI Automation ### Model-Agnostic Enterprise Gateway with Local Open-Source Fallback Intermediate~30 min **Why it's worth it:** Prevents enterprise vendor lock-in, bypasses API rate limits, and reduces operational downtime through local model failovers. Deploys a Python-based abstraction gateway that routes requests to proprietary APIs like Claude 3.5 Sonnet by default, but automatically reroutes traffic to a locally hosted open-source model (e.g., Mistral-7B via vLLM) if the primary API fails or hits rate limits. Python 3.10+Anthropic SDKvLLMRequests 1. Set up a clean Python virtual environment and install primary client libraries on macOS or Linux. python3 -m venv ai_gateway && source ai_gateway/bin/activate pip install anthropic vllm requests 2. Set the required API environment variables in your terminal environment. export ANTHROPIC_API_KEY="your_api_key_here" 3. Spin up a local vLLM model server hosting an open-source model instance to act as a local fallback cluster. python3 -m vllm.entrypoints.openai.api_server --model mistralai/Mistral-7B-Instruct-v0.2 --port 8000 4. Deploy and execute the hybrid gateway router script (gateway.py) to manage API requests dynamically. import os import requests from anthropic import Anthropic def route_prompt(user_prompt: str) -> str: try: client = Anthropic(api_key=os.environ.get("ANTHROPIC_API_KEY")) response = client.messages.create( model="claude-3-5-sonnet-20241022", max_tokens=1024, messages=[{"role": "user", "content": user_prompt}] ) return response.content[0].text except Exception as err: print(f"[GATEWAY WARN] Proprietary API failed ({err}). Rerouting to local vLLM cluster...") payload = { "model": "mistralai/Mistral-7B-Instruct-v0.2", "messages": [{"role": "user", "content": user_prompt}], "max_tokens": 1024 } fallback_res = requests.post("http://localhost:8000/v1/chat/completions", json=payload) return fallback_res.json()["choices"][0]["message"]["content"] if __name__ == "__main__": print(route_prompt("Summarize enterprise model adoption trends.")) **Links:** https://github.com/anthropic-ai/anthropic-sdk-python · https://github.com/vllm-project/vllm **Sources:** AI News & Strategy Daily | Nate B Jones ### Automated Graph-Based Referral Mapping Pipeline Intermediate1-2 hrs **Why it's worth it:** Bypasses competitive public job posting boards by automatically mapping second-degree internal network referrals. Constructs an automated relationship graph in Neo4j using LLM-extracted relationship triples from connection lists, enabling direct graph traversal queries to locate high-trust referral routes to targeted hiring managers. Python 3.10Neo4j ContainerLangChainOpenAI APIpy2neo 1. Deploy a headless Neo4j graph database container on a local or cloud Docker host. docker run -d --name neo4j-referrals -p 7474:7474 -p 7687:7687 -e NEO4J_AUTH=neo4j/SecretPassword123 neo4j:latest 2. Install standard Python driver dependencies and orchestrator frameworks. pip install neo4j langchain-openai python-dotenv pandas 3. Initialize database connection constraints using the Python Neo4j driver. from neo4j import GraphDatabase driver = GraphDatabase.driver("bolt://localhost:7687", auth=("neo4j", "SecretPassword123")) with driver.session() as session: session.run("CREATE CONSTRAINT FOR (p:Person) REQUIRE p.id IS UNIQUE;") session.run("CREATE CONSTRAINT FOR (c:Company) REQUIRE c.name IS UNIQUE;") 4. Execute a Cypher query to retrieve optimal 2-hop paths connecting a candidate to target hiring managers. MATCH path = shortestPath((me:Person {id: "candidate_01"})-[*..2]-(target:Person {title: "Hiring Manager"})) RETURN path; **Links:** https://github.com/neo4j/neo4j-python-driver · https://python.langchain.com/docs/get_started/introduction **Sources:** Joshua Fluke ### Multi-Agent Collision Guard for Parallel Code Editing Advanced~45 min **Why it's worth it:** Prevents file overwrites and race conditions when running multiple autonomous AI coding sessions on a single repository. Sets up a system hook mod inside Claude Code that monitors file modification metadata. When multiple background agent sessions attempt to write to the same file path within a specific lookback window, execution pauses and alerts the user. Claude Code Desktop AppClaude Code CLINode.js Hooks 1. Instruct Claude Code to build a local repository hook that intercepts file system modification commands. Create a local system hook named 'Collision Guard' that intercepts write, replace, and edit operations across active agent sessions. 2. Configure the inspection logic to check recent file write timestamps before applying edits. Set a lookback threshold of 30 minutes. Inspect target file metadata prior to write execution. If the file was modified by a parallel session, halt execution and prompt for user confirmation. 3. Verify hook protection by launching two concurrent Claude Code terminal instances attempting edits on the same file. claude --project ./src/index.js **Sources:** Nate Herk | AI Automation ### LLM-as-a-Judge Optimization Loop for Quality Evaluation Intermediate1-2 hrs **Why it's worth it:** Automates quality evaluation for non-binary, subjective tasks like customer support responses or content moderation. Establishes a programmatic alignment loop that compares an evaluator LLM's scores against a human-graded reference dataset. The evaluator prompt is systematically refined until LLM scoring concordance with human judgment exceeds 90%. Claude Opus 5.5GPT-6 AstraPythonSpreadsheets 1. Collect a benchmark set of ~100 real user queries and manually grade generated agent outputs in a spreadsheet. # Sample dataset format: prompt, model_output, human_score (1-5), feedback_reason 2. Pass the benchmark inputs and outputs through an initial evaluator model system prompt. evaluator_prompt = "You are an expert evaluator. Grade this response on a 1-5 scale for accuracy, brand voice, and clarity." 3. Measure baseline agreement between human grades and evaluator grades, then iteratively feed misaligned outputs back into the system prompt generator. python3 align_evaluator.py --dataset benchmark_100.json --threshold 0.90 **Sources:** Nate Herk | AI Automation ### 24/7 Autonomous Market Scanner with Computer-Use Agents Beginner~10 min **Why it's worth it:** Saves hours of manual web search by executing continuous, background browser monitoring tasks automatically. Deploys ChatGPT Dots in an ambient, continuous execution mode to navigate target websites via browser control, evaluate dynamic listings against explicit search parameters, and report updates on a scheduled recurring loop. ChatGPT DotsOpenAI Computer-Use ModelChatGPT Scheduled Tasks 1. Open the unified ambient Dots interface located at the top of your workspace dashboard. # Access ChatGPT Dots interface via web app or desktop overlay 2. Direct the model with specific filtering parameters and instruct it to execute continuous background scans. Scan local housing market websites for rental listings between $3,000 and $6,000 per month with at least 3 bedrooms. Re-check every 30 minutes continuously and notify me when new matches appear. 3. Inspect agent execution visually in the automated browser side-panel, refining parameters using voice or text commands as needed. Update scanner parameters to require an attached garage and private backyard. **Sources:** The AI Advantage ## Videos Covered Today * Joshua Fluke — WHY JOB SEARCH ADVICE IS COMPLETELY USELESS NOW * AI News & Strategy Daily | Nate B Jones — Anthropic's $500 billion data center bet #ai * AI News & Strategy Daily | Nate B Jones — Microsoft Compared OpenClaw To A Virus. Now It's Bringing It To Your Employer As Autopilot. * Nate Herk | AI Automation — Claude Code Mods Are Game Changers. This One Saves Me Money. * Nate Herk | AI Automation — Claude Code Mods Are Game Changers. Set Up These 5 NOW. * Nate Herk | AI Automation — How to Actually Build & Sell Software with AI as a Non-Techie * Nate Herk | AI Automation — I Tested Codex's NEW $500/mo Ultrafast mode * Nate Herk | AI Automation — I Tested Opus 5.5 vs Sonnet 5.5 * David Shapiro — I can't do everything * David Shapiro — I hate to admit this * The AI Advantage — How ChatGPT Dots Can Change Your Life in 4 Minutes Generated and deployed by Hiro Digest Engine v2.3.8

Repeated strikes on Kyiv’s Southern Bridge signal a new phase in Russia’s drone campaign – one aimed at disrupting the capital’s daily movement.

The crew of a corporate-owned space vessel must stop a dangerous shape-shifting organism after it gets on board. Sounds very familiar.

The generative AI landscape is constantly shifting, and Modelers.cn—Huawei’s Ascend AI ecosystem platform—has just introduced a fresh tool for developers to explore. We are looking at the…

Global automaker Stellantis has announced a three-week shutdown for its Windsor, Ontario plant, starting later this month, due to tariffs and shifting consumer demand in North America. The company stated that the Windsor Assembly Plant will be idle during the weeks of October 19, 26, and November 2, following a two-week shutdown earlier this month. Over 6,400 employees at the facility were laid off for the previous two weeks and are scheduled to return to work on Monday. Stellantis cited 'evolving market conditions, including consumer demand and the impact of tariffs' as the primary reasons for the operational adjustments. The announcement comes amid ongoing contract negotiations between Stellantis and Unifor, the workers' union, which have reached a standstill. The impasse centers on the future of Stellantis' Brampton, Ontario plant, which the company has indicated it wants to sell due to 'market and trade conditions.' Unifor has criticized this plan as 'indefensible.' The union's Local 444 President, James Stewart, expressed concern over the additional downtime, stating it was 'something we didn't want to hear.' The collective agreement between Unifor and Stellantis expired on September 20, and workers are currently following the terms of the expired agreement. Stewart mentioned that the union is preparing a leaflet to update workers on the bargaining process and future expectations. The Windsor plant produces Chrysler Pacifica minivans and Dodge Charger muscle cars, and the company's casting plant in Etobicoke is also part of the negotiations. The situation reflects broader challenges in the automotive industry, with U.S. tariffs impacting Canadian manufacturers and trade relations between Canada and the United States. Stellantis' decision to idle the Windsor plant highlights the company's strategic response to market pressures, while the unresolved contract talks underscore the tension between management and labor over the future of manufacturing operations in Canada.

Luke Travers made a crucial defensive play during a recent sporting event by intercepting a poorly executed pass from Jaylon Brown. The incident occurred during a match on October 3, 2026, where Brown's careless handling of the ball led to a turnover that was capitalized on by Travers. The interception was described as a 'shocker' by commentators, highlighting the unexpected nature of the play. This moment significantly impacted the game's momentum, shifting the advantage to Travers' team. The event was widely covered by sports media, with highlights being shared across various platforms. The match itself was part of a regular season fixture, though specific details about the teams involved were not provided in the available content. The interception demonstrated Travers' sharp defensive skills and ability to read the play, which has been noted as a key factor in his team's performance this season. The incident has been discussed in post-game analysis, with coaches emphasizing the importance of ball security in high-pressure situations. Fans and analysts alike have praised Travers for his quick reaction and positioning, which prevented a potential scoring opportunity for the opposing team. The play has become a talking point in the sports community, with many highlighting it as a defining moment in the game. The match was played in front of a live audience, with the interception occurring in the latter stages of the contest. The significance of this play has been underscored by the fact that it was featured in the 'Latest in Videos' section of the news platform, indicating its relevance to the broader sports narrative. The event has also been referenced in discussions about defensive strategies in the sport, with experts suggesting that such plays can often determine the outcome of closely contested matches. The interception was not just a moment of individual brilliance but also a testament to the team's overall defensive coordination. Travers' ability to anticipate and react to Brown's mistake was crucial in maintaining his team's lead. The play has been analyzed in detail by sports commentators, who have noted the precision required to make such an interception in a fast-paced game. This moment has been added to the growing list of memorable defensive plays in the sport's recent history. The match was part of a larger season where both teams were competing for playoff positions, making the interception particularly significant. The incident has been widely shared on social media platforms, with fans celebrating Travers' contribution to the team's success. The play has also been used as a teaching example in coaching sessions, emphasizing the importance of defensive awareness and quick decision-making. The interception has been recognized as a pivotal moment that shifted the game's dynamics, ultimately contributing to the team's victory. The event has been covered extensively by the sports media, with multiple angles and replays being broadcast to audiences worldwide. The significance of this play has been acknowledged by both fans and professionals in the sport, highlighting its impact on the game's outcome.

Create an account or log in to Instagram - Share what you're into with the people who get you.

Most of the time, the scene outside the Immigration and Customs Enforcement (ICE) facility in the small, coastal town of Scarborough, Maine, could be confu

Cybersecurity teams are shifting toward continuous control across identity, cloud, endpoints, telemetry, and human risk.

Click to read full live report and analysis.

Cybersecurity teams are shifting toward continuous control across identity, cloud, endpoints, telemetry, and human risk.

Click to read full live report and analysis.

Featuring: Cybersecurity is being reshaped by the expansion of cloud infrastructure, AI, distributed systems, and increasingly complex digital environments. As organizations manage more identities, devices, data, and internet-facing infrastructure, security is shifting toward continuous visibility, control, and the ability to respond to risk at scale. This report examines how core areas of