Good Morning from San Francisco,

GPT-5 leaks multiply as OpenAI preps for August launch. One problem: they can't handle current demand.

Google abandons search chaos, sorting results with AI instead. Publishers hate it.

Tesla trades cars for robots as sales crater. Musk's politics killed his customer base. Now he promises robotaxis everywhere while struggling to get permits anywhere.

Three companies. Three desperate pivots. Same question: Can AI promises save failing businesses?

Stay curious,

Marcus Schuler

GPT-5 Leaks Point to August Debut But Infrastructure Questions Remain

GPT-5 rumors are gaining steam. The Verge reports OpenAI targets an August launch, with Microsoft engineers prepping servers for what could be the biggest AI release since ChatGPT launched.

The evidence keeps coming. Engineer Tibor Blaho found a config file mentioning "GPT-5 Reasoning Alpha" dated July 13. Researchers spotted GPT-5 references in OpenAI's internal biosecurity tools. Even OpenAI's Xikun Zhang casually said GPT-5 "is coming" during a recent discussion.

GPT-5 aims to unify everything into one system. Instead of switching between vision, reasoning, and task completion modes, users get one model that handles it all. Three versions will launch: main, mini, and nano.

But there's a gap between vision and reality. Sam Altman wants free GPT-5 for everyone on Earth - all 8 billion people. OpenAI still struggles with current demand and had to limit GPT-4.5 access after running out of GPUs.

The stakes go beyond user experience. If GPT-5 achieves artificial general intelligence, Microsoft loses revenue rights under their partnership deal. Safety researchers worry about GPT-5's biosecurity testing, questioning whether safeguards can prevent misuse.

OpenAI plans to release an open reasoning model before August to test capabilities and gather feedback.

Why this matters:

• The leaks suggest GPT-5 is closer than OpenAI admits, but the company's grand promises still clash with basic capacity problems that plague current models.

GPT-5 August Launch Rumors Heat Up Despite Capacity Issues

OpenAI’s GPT-5 launch rumors heat up with August timeline leaks, but there’s a catch: Altman promises free AI for 8 billion people while the company still can’t handle current ChatGPT demand. The gap between vision and reality keeps growing.

implicator.aiMarcus Schuler

AI Image of the Day

Prompt:
brown rat in a roman gladiator costume, like Russel Crowe, cinematic, 8k

Google's Web Guide puts search results in neat little boxes

Google launched Web Guide on Thursday. The new feature uses AI to sort search results into categories like "safety tips" and "personal experiences." Think digital filing cabinets for your search page.

The tool runs on Gemini AI through Google's Search Labs program. You can opt in or stick with regular search chaos. Web Guide breaks queries like "solo travel in Japan" into themed sections - travel guides, safety advice, personal stories.

Behind the scenes, Google uses "query fanout technique." The system runs multiple related searches at once, then groups results by topic. Complex questions get split into parts with separate categories for each aspect.

The feature starts in the Web tab without AI Overviews cluttering the page. You see actual links at the top instead of AI-generated summaries. Google plans to expand it to the main "All" tab later.

Publishers aren't thrilled. Pew Research found only 8% of people click links when AI summaries appear, compared to 15% without them. The News/Media Alliance already called Google's AI Mode "theft."

Google faces antitrust pressure while competing with ChatGPT. Web Guide represents a middle path - using AI to organize results rather than replace them entirely.

Why this matters:

• Google retreats from aggressive AI integration after publisher backlash but still wants control over how people discover information.

• This could determine whether the open web survives or gets replaced by platform-controlled content systems.

Google’s Web Guide Sorts Search Results with AI

Google’s Web Guide sorts search results with AI after backlash over AI Overviews. Publishers fear traffic theft as Google seeks middle ground between traditional search and AI dominance. Could reshape how we discover information.

implicator.aiRobert Brown

🧰 AI Toolbox

How to Build AI Customer Service Bots

Quickchat AI creates chatbots that talk like real people. Train the bot with your company information and it answers customer questions naturally. Works on websites, messaging apps, and phone systems.

Tutorial:

1. Sign up for a Quickchat AI account
2. Create a new AI assistant and name it
3. Upload your company information to train the bot
4. Set the bot's personality and conversation style
5. Add the bot to your website or messaging apps
6. The bot starts helping customers immediately
7. Monitor conversations and take over when needed

URL: https://www.quickchat.ai/

Better Prompting

Today: Balanced Scorecard Devbelopment Guide

I need to create a balanced scorecard for my [type of organization/department]. Please provide a step-by-step guide that covers:

1. How to identify and define objectives for each of the four perspectives (Financial, Customer, Internal Process, Learning & Growth)
2. How to develop meaningful KPIs and set realistic targets
3. Best practices for aligning metrics with our strategic goals
4. Common pitfalls to avoid during implementation
5. A simple template or framework I can start with

Background: [Brief context about your organization size, industry, or specific challenges] Experience level: [Beginner/Intermediate/Advanced] with performance management tools

AI & Tech News

Intel Cancels Germany and Poland Factories Despite Revenue Beat

Intel's new CEO Lip-Bu Tan canceled planned chip factories in Germany and Poland while the company posted better-than-expected revenue of $12.86 billion in the second quarter. The stock still dropped 5% after hours as Tan declared "no more blank checks" and cut the workforce to 75,000 employees, admitting Intel "invested too much, too soon" without adequate demand for its foundry business that lost $3.17 billion.

Microsoft CEO Calls Mass Layoffs an "Enigma" While Spending $80 Billion on AI

Microsoft CEO Satya Nadella sent a company-wide memo acknowledging the "enigma" of laying off over 15,000 employees this year while posting record profits and spending $80 billion on AI infrastructure. The memo came as internal unrest grows over what employees see as a return to the "old Microsoft" culture of fear and job insecurity, with 9,000 workers cut in July alone despite the company's claims of thriving by "every objective measure."

Apple's macOS Tahoe Beta Brings Powerful Search Updates and Half-Baked Design

Apple released the macOS Tahoe public beta with major upgrades to Spotlight that add a built-in clipboard manager, app actions, and four search modes, plus Shortcuts automations that can trigger based on folder changes and AI integration. The new "Liquid Glass" design feels unfinished and messy on Mac, with toolbars that look like "flat light gray ovals separated from a featureless gray expanse by an amateurish drop shadow," but the productivity gains from Spotlight and Shortcuts improvements outweigh the visual quirks.

Starlink Goes Dark for 2.5 Hours in First Major Outage of 2025

Starlink's satellite internet went down for 2.5 hours Thursday afternoon due to failed internal software, marking the first major outage in 2025 and dropping global connectivity to just 16% of normal levels. The blackout hit Ukrainian troops particularly hard, cutting off communications across the entire front line for what the military called "the longest outage in the war."

X Searches for Rare Species: Posts That Unite People

X is testing a new feature that uses its Community Notes system to highlight posts liked by users with different viewpoints, applying the same "bridging algorithm" that finds consensus among people who typically disagree. The experiment aims to surface content that "resonates broadly" on a platform better known for dividing people than bringing them together.

Walmart Consolidates Confusing AI Tools Into Four Simple Interfaces

Walmart built dozens of AI agents but found users got confused switching between different interfaces, so now it's consolidating everything into four "super agents" for customers, employees, engineers, and suppliers. The customer-facing agent called Sparky is already live while the supplier agent Marty launches soon, with Walmart using Anthropic's Model Context Protocol to connect all the behind-the-scenes tools into unified experiences.

China's Booming Black Market: Fixing Banned Nvidia AI Chips

A dozen Chinese repair shops now fix up to 500 banned Nvidia AI chips per month, charging $1,400-$2,800 per repair and proving that large-scale smuggling continues despite U.S. export restrictions. The thriving repair industry exists because smuggled H100 and A100 chips have been running 24/7 in Chinese data centers for years and are now failing at higher rates, with some companies even building 256-server facilities to test repairs.

Aidoc Lands $150M from Nvidia to Speed Up Medical AI Development

Israeli medical AI company Aidoc raised $150 million with Nvidia's investment arm joining for the first time, bringing total funding to $370 million as the company's new foundation model builds algorithms ten times faster than before. The funding will accelerate development of AI tools that help doctors spot life-threatening conditions in real-time across thousands of hospitals worldwide, with revenue from the new technology already exceeding everything the nine-year-old company made in recent years.

Former Carpenter Raises $40M to Fix Construction's Paper Problem

Trunk Tools raised $40 million in Series B funding to build AI that makes sense of construction's endless documents, drawings, and specs that bog down the $11.4 trillion industry. The company was founded by Dr. Sarah Buchner, who started as a carpenter at age 12 in Austria before earning a PhD and deciding the construction world needed better than "passive digital repositories" for managing project data.

Musk's Politics Crush Tesla While He Chases Robotaxis

Tesla's crisis keeps getting worse. The company reported its third straight quarter of falling profits, with revenue down 12% to $22.5 billion. Car sales dropped 16% while Musk warned of "rough quarters" ahead.

Musk's politics broke his business. His $290 million Trump backing and far-right activism sparked global protests. Liberal buyers—Tesla's original customers—walked away. In Europe, Tesla's market share fell for six straight months to 2.8%.

The policy hits keep coming. Trump's bill kills the $7,500 EV tax credit on September 30. Tesla also loses regulatory credit sales that brought in $890 million last year. Tariff costs jumped $300 million this quarter alone.

Now Musk bets everything on robotaxis. He promises to reach half the U.S. by year-end, but Tesla hasn't even applied for California permits. Waymo already runs driverless cars in multiple cities while Tesla's tiny Austin fleet needs human supervisors.

Tesla trades at $1 trillion on robot promises. Strip away the hype? It's another struggling car company with falling sales and old models.

Why this matters:

• Musk turned Tesla's core customers into enemies through toxic politics, showing how CEOs can destroy billion-dollar brands overnight

• The robotaxi pivot looks desperate as regulatory reality clashes with promises—if this fails, the trillion-dollar valuation becomes impossible to justify

Tesla’s Crisis: Profits Fall as Musk Bets on Robotaxis

Tesla reports third straight quarter of falling profits while Musk’s political activism alienates core customers. Now the company bets everything on unproven robotaxis as EV sales plummet and regulations pile up.

implicator.aiMaria Garcia

🚀 AI Profiles: The Companies Defining Tomorrow

Boson AI 🎯

Two Amazon AI vets ditched the cloud giant to teach machines how to actually talk. They're burning through a tiny war chest while giants circle overhead.

• The Founders
Dr. Alex Smola and Dr. Mu Li launched Boson AI in 2023 after years building AWS's ML backbone. The 30-person team camps in Santa Clara, with a Toronto outpost hunting Canadian AI talent. Mission: make chatting with AI feel human instead of robotic.

• The Product
Higgs Audio devours 10 million hours of training data to birth emotionally intelligent voices. Their LLM-powered speech engine doesn't just read text—it gets context, switching between character voices mid-conversation and injecting actual personality. The flip side? Audio understanding that catches tone, subtext, and frustration levels.

• The Competition
ElevenLabs already snagged unicorn status at $3B. Character.AI hit $1B for text-based companions. Inworld AI grabbed $500M for gaming NPCs. Meanwhile, OpenAI gives ChatGPT a voice while Amazon sharpens Alexa's brain. It's David vs. an army of Goliaths. 💀

• Financing
Ouch. Just $1.2M in seed money versus competitors swimming in hundreds of millions. Valuation likely sits in single-digit millions—pocket change in today's AI gold rush. The team's pedigree opens VC doors, but they need serious cash soon.

• The Future ⭐⭐⭐
Strong tech foundation meets brutal funding disadvantage. Open-source strategy could build community momentum, but scaling voice AI burns serious compute dollars. Either they land enterprise customers fast or become acquisition bait for a tech giant. The race is on. ⏰

Bottom line: World-class engineers with champagne taste on a beer budget, racing against time and titans.