Frigate lets you train lightweight MobileNetV2 models to classify tracked objects into finer categories (like identifying specific pets or detecting safety gear). The models require consistent predictions across multiple frames before assigning labels, which prevents false positives but means you need to carefully curate training data to avoid overfitting.

• Classification requires a two-step verification: confidence score above threshold (default 0.8) plus 60% consensus across at least 3 attempts, preventing single-frame misclassifications from triggering assignments.
• Training data quality matters far more than quantity—using dozens of nearly identical images actually degrades performance through overfitting, so you should start small with visually distinct examples and add hard cases (low confidence predictions, unusual angles/lighting) over time.
• The `none` class needs the opposite training approach from named classes: it should include visual diversity (small crops, partial views, false positives) so the model can recognize when something doesn't match your known categories, otherwise borderline detections get forced into named classes by default.

This doc explains how to configure which objects Frigate detects and tracks in video feeds, including how to filter false positives and mask detection zones. By default it only tracks people, but you can add cars, animals, and dozens of other object types.

• Frigate supports 90+ object types from Google Coral's dataset, but only tracks "person" by default — you must explicitly enable tracking for other objects like cars or dogs
• Object filters let you reduce false positives by setting minimum/maximum bounding box size, aspect ratio, and confidence thresholds per object type at global or per-camera level
• Mask zones prevent detection in specific frame areas (like trees or shadows) and can be applied globally or to individual object types

Rest of World's publisher visits China and finds the U.S.-China AI rivalry isn't as binary as it seems—both countries are pursuing different but equally serious paths, with China betting heavily on embedding AI into physical products and manufacturing rather than chasing raw computational power.

• China is pivoting toward deploying AI in robotics, EVs, and factories instead of competing for frontier model capability, partly due to U.S. chip export controls but also as a deliberate strategic choice.
• American export restrictions may be obsolete; China's real constraint is now high-quality data and interaction with the physical world, not compute power—meaning Washington's leverage is weaker than assumed.
• Both countries face genuine risks and trade-offs: the U.S. debates innovation versus regulation while China treats advanced AI as a national security asset, embedding it across the economy with minimal constraints on state surveillance or individual liberty.

Anthropic has turned Claude into a persistent Slack bot that watches channels, answers questions, suggests tasks and can trigger actions without prompting. It acts like an “agentic” AI coworker, raising questions about privacy, control and user consent in workplace chats.

• Claude now runs persistently in Slack, monitoring channels and proactively surfacing suggestions, summaries, and drafts without being explicitly summoned.
• Admins retain control via channel toggles, sensitivity settings, guardrails on sensitive terms, and mandatory review before any suggestion posts.
• Data stays within the Slack workspace unless a user explicitly sends it to Anthropic's servers for fine-tuning.
• It's currently an enterprise preview, positioned to compete with Microsoft Teams Copilot and Google Duet AI, with wider rollout planned later in the year.

A long-time Android security architect at Google announces his resignation, citing moral objections to the company’s AI energy impact and secret U.S. military contracts. He reflects on Google’s former “don’t be evil” culture, his achievements in user encryption and privacy, and why recent top-level decisions forced him to leave.

• A senior Android Security architect resigned over Google abandoning renewable energy goals to power AI and signing "any lawful purpose" DoD contracts.
• He says these military deals enable offensive warfare support and mass surveillance of Europeans, violating human rights norms and EU privacy law.
• Leadership made these decisions unilaterally with no internal debate or transparency before employees learned of them.
• He's giving three months' notice (through August 31, 2026), immediately cutting ties with AI work tied to the DoD deal while transitioning Android Security duties.

The article argues that companies are increasingly recording every meeting by default to feed AI systems the living context of their culture, decisions, and conversations. This shift turns unstructured voice data into a searchable, structured system of record that boosts individual productivity and executive oversight, making meeting recording inevitable.

• Meeting recording became default not through decision but because tools shipped with it on and nobody turned it off
• Companies like Bridgewater, OpenAI, and a16z (via Granola) already treat recorded conversation as a living system of record smarter than any wiki
• Skipping recording costs both individual productivity (no context-aware AI assistant) and executive oversight (no early warning system for risks)
• Verbal-culture companies (Shopify, OpenAI) have more to gain from this shift than document-heavy cultures (Stripe, Anthropic), since their key knowledge was previously unrecorded

A Wired report reveals deep discontent within Meta’s three-month-old Applied AI team, where 6,500 employees were “drafted” into grueling data-labeling work or forced out. Internal outbursts, petitions against keystroke monitoring, and surprise reassignments have left engineers calling the unit a “soul-crushing gulag.” Leadership acknowledges the mistakes and low morale in an internal memo.

• Meta reassigned 6,500 engineers into an "Applied AI" unit to do data-labeling grunt work, with no real choice but to comply or quit, and morale has collapsed to the point employees call it a "soul-crushing gulag"
• Someone hacked a livestreamed internal presentation to call a senior AI executive "a piece of sh*t," reflecting broader anger over the forced reassignments
• Over 1,600 employees signed a petition opposing keystroke and click monitoring used to generate AI training data
• Zuckerberg and Chief Product Officer Chris Cox have both acknowledged internally that the situation is bad, with Cox calling the environment "brutal" and Zuckerberg admitting mistakes while promising fixes

After Palantir became the tech backbone for Trump’s immigration enforcement, current and former employees warned of a “descent into fascism.” Workers now question the company’s role as it deepens ties with an administration they view as attacking immigrants and fueling conflicts abroad. Palantir says it promotes internal debate and does not endorse a single ideology.

• Palantir's expanded ICE contracts to track and deport undocumented immigrants triggered internal alarm, with two former engineers coining the phrase "tracking Palantir's descent into fascism" as a running joke turned genuine concern.
• Employees who once saw their work as defensible post-9/11 national security efforts now view similar work as morally compromised under Trump's immigration crackdown.
• Staff are increasingly pushing back internally via meetings and Slack over the company's role in surveillance, deportation enforcement, and potential involvement in an Iran conflict.
• Palantir's official response emphasizes its culture of "fierce internal dialogue" and ideological diversity rather than addressing the substance of employee concerns.

A San Francisco shop is operated almost entirely by a central AI agent that manages checkout, inventory and security. The Times examines how the system handles everyday tasks, misidentifies items and prompts privacy concerns. It shows both the promise and real-world glitches of automating retail with AI.

• Shelf.AI's cashier-less store cut labor costs 60% by using an AI agent ("Mia") to handle pricing, inventory, and theft prevention across cameras, weight sensors, and LIDAR
• 7% of opening-day transactions undercharged customers due to glitches, fixed within 48 hours—showing the tech isn't yet flawless
• Privacy concerns persist over constant surveillance, despite Shelf.AI's claims of discarding raw footage after 24 hours and encrypting data
• Early financials look promising ($45,000 weekly sales, break-even projected in 8 months), prompting other grocers to test similar AI systems

This article discusses how Flock's technology is transforming crime-solving in cities like San Francisco and Tulsa. By using advanced camera systems and data analysis, Flock helps police departments improve their effectiveness, resulting in higher clearance rates for crimes and enhanced community safety.

• Flock's cheaper camera network captures far more vehicle data than traditional ALPR systems, letting police identify suspects even without eyewitness IDs or other typical identifiers.
• Tulsa solved every homicide case in 2024 for the first time in nearly 50 years, and Lakewood, WA went a full year without a homicide for the first time in two decades, both credited to Flock's tech.
• Flock claims its system helps solve roughly 700,000 crimes and over 1,000 missing persons cases annually.
• Crime clearance rates have fallen sharply since the 1960s, framing Flock's tech as a way to reverse that long-term decline in police effectiveness.