Impli reveals the exact APEX method professionals use to optimize AI prompts. The article shows the complete system specification—from analyzing requests to executing optimized prompts that work across all platforms.
Web scraping has quietly become the backbone of AI training data. But legal gray areas and sophisticated anti-blocking measures make success tricky. This guide reveals what works in 2025.
Sean Grove from OpenAI says coding is dead. Instead of writing code, developers should write specifications that generate software. AWS just launched Kiro to make this real, while GeneXus claims they've done it for 35 years
OpenAI's AI now controls computers while Trump prepares an order forcing AI companies to prove political neutrality for federal contracts. Tech giants must choose: adjust development or lose billions as China advances without limits.
Good Morning from San Francisco,
OpenAI launched ChatGPT Agent Thursday. First AI that actually controls your computer.
The agent books restaurants, manages calendars, writes reports. Takes 15-30 minutes per task - slower than you'd like, faster than doing it yourself.
This isn't web browsing with training wheels. It manipulates software, accesses terminals, connects APIs. During demos, it planned date nights by matching Google Calendar with OpenTable.
Performance doubled on benchmarks. Pro subscribers get access now.
Stay curious,
Marcus Schuler
OpenAI launched ChatGPT Agent Thursday. It's the first AI that can control your entire computer, not just browse websites.
The agent handles tasks like booking restaurants, managing calendars, and writing research reports. It takes 15-30 minutes per task - slower than you'd like, but faster than doing it yourself. "Even if it takes half an hour, it's quite a big speed-up," says OpenAI's Isa Fulford.
This isn't just web browsing with extra steps. The agent combines OpenAI's previous tools - Operator and Deep Research - into something that can manipulate software, access terminals, and connect to APIs. During demos, it planned date nights by cross-referencing Google Calendar with OpenTable.
The performance jumps are real. ChatGPT Agent scored 41.6% on Humanity's Last Exam, double what OpenAI's other models achieved. On FrontierMath, it hit 27.4% versus 6.3% previously.
OpenAI added safeguards because, well, AI controlling computers sounds terrifying. The agent asks permission before "irreversible" actions like sending emails. Financial transactions are restricted. There's even a Watch Mode that stops working if you navigate away from financial sites.
The tool is available now to Pro, Plus, and Team subscribers. Enterprise users get it this summer. Europe will have to wait.
Why this matters:
• We've moved from AI that talks to AI that acts - this is the first tool that can actually control your computer instead of just chatting about it
• The benchmark improvements suggest we're hitting an inflection point where AI moves from "sounds smart" to "handles real work"
Robert Brown
Prompt:
Close-up of a white-haired old Thai woman on a
dark purple background with decorative light effects. She is wearing a traditional Thai patterned dress, with small white flowers attached to her clothes. The light creates a creepy, creepy dim effect with shades of purple and pink around her hair, making her hair shimmer. This image has a Thai cinematic quality with strong color intensity, especially in the purple tones. The composition is slightly tilted, with the subject slightly turned to the side. The background has an abstract light pattern that creates a bokeh effect, adding depth to the creepy image.
Trump's team is preparing an executive order that would force AI companies to prove their systems are politically neutral if they want federal contracts.
The move targets what officials call "woke AI" after incidents like Google's Gemini showing racially diverse Nazis and depicting George Washington as Black. David Sacks, Trump's AI czar, and Sriram Krishnan are leading the effort with a 20-page plan.
The executive order is part of a three-pronged strategy. The administration also plans to speed up permits for data centers and promote exports of U.S. chips to allies like the United Arab Emirates. The goal is beating China in the AI race while addressing conservative complaints about liberal bias.
Major players like Microsoft, Google, and Anthropic would need to adjust their AI development to secure government contracts worth billions. Anthropic faces particular scrutiny after hiring Biden officials and opposing Trump positions on chip exports.
The tech industry will likely resist political constraints on AI development. Companies argue that forcing neutrality could slow innovation while Chinese firms advance without similar limits.
The three executive orders are expected by month's end, with Trump planning to speak about his AI "action plan" then.
Why this matters:
• The government is using federal contracts worth billions to shape political ideology in AI technology
• Political requirements could slow U.S. AI progress while China races ahead without similar constraints
Marcus Schuler
Scalenut helps you create blog posts and articles that show up in Google search results. It researches what's already ranking, creates an outline, and writes content optimized for search engines.
Tutorial:
URL: https://www.scalenut.com/
Perplexity's valuation shot up to $18 billion just two months after raising money at a $14 billion price tag, as investors scramble for access to hot AI startups. The three-year-old Google search challenger has now raised money five times in 18 months, ballooning from a $500 million valuation at the start of 2024 to today's $18 billion after launching an AI web browser and growing annual revenue from $35 million to $150 million.
Sam Altman transformed from a Democrat who called Trump an "unprecedented threat to America" and compared him to Hitler into the president's AI confidant, outmaneuvering former ally Elon Musk through quiet lobbying and billion-dollar infrastructure deals like Stargate. The OpenAI CEO declared himself "politically homeless" after leaving the Democratic Party, securing Pentagon contracts and influence over AI policy while Musk's relationship with Trump crumbled over disputes about Middle East data center projects.
Netflix used generative AI for visual effects in one of its original shows for the first time, creating a building collapse scene in Argentine sci-fi series "The Eternaut" that was completed 10 times faster than traditional methods. The move comes as the streaming giant reported a 16% revenue jump to $11 billion amid ongoing industry debates about AI replacing human workers in entertainment production.
Google launched its Veo 3 video generation model on the Gemini API, charging $0.75 per second for 720p video with synchronized audio that includes dialog, music, and sound effects from text prompts. The pricing means an eight-second clip costs $6 and a five-minute video runs $225 - though developers often need multiple attempts to get usable footage, potentially driving costs much higher.
Meta hired Mark Lee and Tom Gunter, two key Apple AI researchers who worked under Ruoming Pang - the Apple executive Meta poached earlier this month with a compensation package worth over $200 million. The moves reflect Meta's aggressive talent grab in AI, with the company offering packages worth several multiples of Apple's pay while Apple's own AI team faces uncertainty about whether to use outside models for Siri.
Anthropic tightened usage limits for Claude Code since Monday without telling users, leaving many $200-a-month Max plan subscribers unable to work after hitting unexplained restrictions within hours. The company confirmed the issues but declined to explain the changes, frustrating users who say they can't plan projects around invisible limits that reset at random times.
A coalition including the Gates Foundation and Charles Koch's Stand Together will spend $1 billion over 15 years to develop AI tools for public defenders, parole officers, and social workers through a new entity called NextLadder Ventures. The group partnered with Anthropic to create software that helps frontline workers manage massive caseloads, from matching disaster victims with resources to processing parole paperwork faster.
Mistral launched a "deep research" mode for its Le Chat chatbot that can plan, search, and synthesize data from web sources, bringing the French AI lab closer to competing with OpenAI and Google's offerings. The company's key advantage lies in connecting to enterprise data on-premises rather than through cloud services, letting banks, defense contractors, and government agencies use AI tools on sensitive data without uploading anything to external servers.
Meta won't sign the European Union's voluntary AI code of practice, with the company's Joel Kaplan declaring that "Europe is heading down the wrong path on AI" and calling the guidelines overreach that go beyond the bloc's AI Act. The snub comes as US tech companies continue their standoff with European regulators, with Trump's administration having lobbied against the code and dozens of European firms asking for a two-year timeout on AI Act implementation.
Russian lawmakers warned WhatsApp to prepare for an exit from the country, calling the Meta-owned messaging app a national security threat that will likely face restrictions under Putin's September 1 deadline for limiting "unfriendly" software. The push comes as Russia promotes its homegrown MAX messaging app to replace WhatsApp, which currently serves 68% of Russians daily, in the country's broader effort to achieve what it calls digital sovereignty.
Firestorm Labs raised $47 million in Series A funding to expand its portable xCell factories that can 3D print combat drones anywhere in the world within two hours of setup. The California company, which has Defense Department contracts worth up to $100 million, plans to scale production of the weapons and tools defining the Russia-Ukraine war while securing exclusive distribution rights for HP's advanced 3D printing technology.
Create a comprehensive talent acquisition and retention strategy for a [industry/sector] startup in the [early/growth/scale] stage with [X] current employees. Focus on attracting [specific roles: e.g., software engineers, sales professionals, product managers] within a [budget range/constraints].
Please include:
Context: [Brief description of your startup's mission, current team size, funding stage, and any specific challenges you're facing in hiring]
Deliverable: A actionable 2-3 page strategic plan with specific tactics, timelines, and measurable outcomes.
Firestorm Labs prints drones where bullets fly. The San Diego startup drops shipping containers packed with 3D printers into combat zones, churning out aircraft in hours instead of months.
• The Founders
Founded 2022 by Dan Magy (serial defense entrepreneur who sold Citadel Defense for nine figures), Chad McCoy (23-year Air Force special ops veteran), and Ian Muceus (aerospace engineer CTO). Based in San Diego near Marine Corps Air Station Miramar. Growing team of engineers across multiple facilities. 🚁
• The Product
xCell mobile factories fit in shipping containers, unfold into battlefield workshops. Tempest drones (7-foot wingspan, 55 pounds) print in 9 hours, snap together by hand. El Niño handheld strike drones weigh under 10 pounds. Modular payloads swap in minutes. One xCell cranks out 50 drones monthly. No GPS required – pure autonomous targeting.
• The Competition
RapidFlight mirrors their model but crashed during head-to-head demos (literally). Currently locked in patent litigation. Legacy contractors like Lockheed invest rather than compete directly. Ukraine's garage innovators prove the concept works. Race is on for Pentagon's attritable drone doctrine.
• Financing
$47M Series A led by NEA, plus $12M venture debt from JP Morgan. Total funding approaches $60M. Lockheed Martin Ventures backs them. Landed $100M Air Force contract spanning 3-5 years. Valuation likely nine figures but undisclosed. 💰
• The Future ⭐⭐⭐⭐⭐
Perfect timing as Pentagon shifts toward "cheap and plentiful" over "expensive and exquisite." NATO allies want resilient supply chains. Ukraine proved distributed manufacturing wins wars. Risk: Pentagon bureaucracy kills nimble companies, but Firestorm's already inside the wire.
Good Morning from San Francisco, TSMC posted a record $13.5 billion profit. AI demand drove the surge. The chipmaker
Good Morning from San Francisco, China spent $6 million to build what cost OpenAI hundreds of millions. 🤖 DeepSeek's
Trump banned Nvidia's China chip sales in April over security fears. Three months later, he reversed it after meeting CEO Jensen Huang. The $15 billion flip shows how economic pressure can override national security concerns in tech policy.
Third 'technical glitch' in six months. The pattern raises questions as Grok launches in Tesla vehicles.
Get the 5-minute Silicon Valley AI briefing, every weekday morning — free.
