PLUS: Google takes on Nvidia, the case for junior devs, and AI's hidden consumption power
Good morning
Google just completed its Gemini 3 family with the release of Flash, a new model that delivers performance nearly on par with its Pro-level AI. This release makes powerful AI more accessible through dramatic cuts in cost and speed.
By making high-performance AI so much cheaper and faster, Google is putting more power directly into the hands of builders. This raises a key question: how will this shift in accessibility reshape the competitive landscape for AI startups and established players alike?
In today’s Next in AI:
Google's new Gemini Flash model
Instacart's AI Price Probe
The case for investing in junior devs
Unlocking insights from personal data
Google's New Flash
Next in AI: Google just released Gemini 3 Flash, its latest and most efficient model that delivers near-Pro level performance at a fraction of the cost and speed. This release completes the Gemini 3 family, making powerful AI more accessible for developers and everyday users.
Decoded:
The new model shows significant improvement in advanced reasoning, tripling the score of its predecessor on the Humanity's Last Exam (HLE) benchmark.
For developers, Flash runs workloads three times faster than Gemini 2.5 Pro and is substantially cheaper, with input tokens costing 75% less than the current Gemini 3 Pro model.
It's already the new default model in the Gemini app and is being rolled out to power Search's AI Mode, providing a notable upgrade for free users.
Why It Matters: Google is aggressively closing the gap between its top-tier and cost-effective models, putting more power into the hands of builders. This move pressures the market by lowering the barrier to entry for creating high-performance, scalable AI applications.
Instacart's AI Price Probe
Next in AI: The U.S. Federal Trade Commission is investigating Instacart over its use of AI-powered pricing tools. The probe follows a study suggesting the practice could cost some shoppers over $1,200 extra per year.
Decoded:
A recent study found prices for identical grocery items could vary by around 7%, highlighting how AI can create invisible price differences for consumers.
Instacart refutes the claims, stating its retail partners control pricing and that its system uses randomized A/B testing, not dynamic pricing based on individual user data.
The company's focus on this technology isn't new; Instacart acquired AI pricing firm Eversight in 2022 to create "compelling savings opportunities," according to a regulatory filing.
Why It Matters: This high-profile investigation signals growing regulatory focus on the opaque ways companies use algorithms to set prices. It serves as a critical test case on corporate responsibility and transparency in the AI-driven consumer economy.
AWS Chief Defends Junior Devs
Next in AI: AWS CEO Matt Garman pushes back on the trend of replacing junior developers with AI, calling it one of the 'dumbest ideas' he's ever heard and a threat to long-term company health.
Decoded:
Junior talent is often more adept with new AI tools, with a survey showing that 55.5% of early-career developers use AI tools daily in their workflow.
Targeting entry-level roles for cost savings is an ineffective strategy, as junior employees are typically the least expensive part of the payroll.
Removing junior roles breaks the talent pipeline, cutting off the supply of future leaders and leaving companies vulnerable to a talent shortage.
Why It Matters: Garman’s perspective is a crucial reminder to balance short-term efficiency gains with long-term strategic health. Building a sustainable workforce means investing in new talent, not just new technology.
AI's Real Superpower
Next in AI: A growing number of developers are finding AI's true strength isn't just generating content, but consuming your personal knowledge base to surface forgotten insights. This effectively turns years of notes into a queryable extension of your own memory.
Decoded:
The key insight is to shift your queries from asking AI to generate new content to asking it to synthesize your own existing knowledge, uncovering patterns you might have missed.
For example, by analyzing years of meeting notes, AI can identify non-obvious correlations, like performance issues consistently preceding complaints about tooling by several weeks.
This method bypasses the limits of keyword search because AI can query by concept, finding thematic connections between ideas you documented years apart.
Why It Matters: This approach transforms your personal archive from a passive repository into an active partner for problem-solving. It effectively lets you tap into the compounded wisdom of your own experience on demand.
AI Pulse
Coursera announced a merger with Udemy in an all-stock transaction, creating a combined company with an implied equity value of approximately $2.5 billion to address the global demand for AI skills.
Micron forecasted surging second-quarter revenue well above estimates, as demand for its high-performance memory chips used in AI data centers continues to boom.
Microsoft confirmed that Windows 11's upcoming AI agents will require explicit user consent before accessing personal files in known folders, addressing recent privacy concerns.