Latest Interviews

GPT-4.5 = Big Model Energy | YC Decoded

Tom Blomfield

OpenAI has released GPT-4.5, a significantly larger and more human-like model that surpasses its predecessor in emotional intelligence and creative fluency while facing higher costs and specific limitations in complex STEM reasoning. Market reception remains measured as benchmarks show incremental rather than revolutionary improvements, leading researchers to predict a future convergence of general and reasoning capabilities in the upcoming GPT-5 architecture. Simultaneously, YC is hosting a free AI Startup School in San Francisco featuring industry leaders such as Elon Musk, Sam Altman, and Satya Nadella to support emerging AI talent.

The Engineering Unlocks Behind DeepSeek | YC Decoded

Diana Hu

DeepSeek unveiled its open-source R1 reasoning model and V3 base architecture, delivering performance parity with top-tier proprietary systems like OpenAI o1 through novel algorithmic optimizations such as native FP8 training and Group Relative Policy Optimization. The announcement triggered significant market volatility, erasing nearly $600 billion from NVIDIA's capitalization, while demonstrating that software stack efficiency can rival the hardware advantages of major US tech ecosystems despite strict GPU export controls. This breakthrough validates a new trajectory for the industry by proving that high-cost AI development can be drastically reduced, thereby lowering barriers for startups and accelerating the adoption of accessible, locally deployable intelligence.

How Scaling Laws Will Determine AI's Future | YC Decoded

Garry Tan

This event analyzes the evolution of AI scaling laws, tracing the transition from the 2020 Chinchilla optimization that prioritized data volume over model size to the current era where OpenAI's O3 model demonstrates that intelligence scales linearly with test-time reasoning compute. The discussion highlights industry concerns regarding pre-training data plateaus while arguing that the scaling revolution is shifting toward reasoning depth and unexplored modalities like robotics and chemical modeling. These insights collectively outline a new trajectory for Artificial General Intelligence that moves beyond simple architectural expansion to leverage extended inference and cross-domain applications.

Why nations that fail women fail

Analysis of global data establishes a direct correlation between patriarchal customs like bride pricing and polygyny and heightened state fragility, as these practices fuel male kinship rivalries and create demographic imbalances that recruit disenfranchised men into armed groups. Despite this evidence, female negotiators currently comprise only 16% of UN-led peace talks, a critical deficit that undermines stability since history shows female-led movements result in lasting peace agreements 35% more often than male-dominated efforts. Policymakers are therefore urged to prioritize women's property rights, enforce minimum marriage ages of 18, and integrate female interests into development aid to transform societal instability into resilient governance.

Was The War Against Japan Avoidable? - Sarah Paine

Sarah Paine

Historian discussions analyze how the Smoot-Hawley Tariff fostered global economic instability that enabled militarist factions in Japan, led by War Minister Tojo, to assassinate pro-peace figures and drive nations into World War II. The dialogue examines the United States' strategic dilemma in responding to the oil embargo, weighing the moral necessity of denying resources to an aggressor against the risk of triggering a broader conflict or inadvertently aiding a potential Nazi victory in Eurasia. Ultimately, the analysis concludes that Japan's modern post-war strength was a contingent outcome rather than an inevitable result, highlighting the extreme difficulty of predicting how specific policy choices might have altered the war's trajectory.

How David Lieb Turned a Failing Startup Into Google Photos | Backstory

David Lieb

David Lieb co-founded the contact-sharing startup Bump, which was eventually acquired by Google in 2015, but his subsequent secret launch of the billion-user Google Photos product resulted in two firings before its 2015 debut. Following a 2020 leukemia diagnosis and a grueling year-long treatment regimen, Lieb transitioned from a product builder to a mentor at Y Combinator. He now leverages his history of strategic pivots and near-death survival to guide and protect the next generation of founders.

Anthropic’s Claude Computer Use Is A Game Changer | YC Decoded

Garry Tan

Anthropic has launched a public beta of "Claude Computer Use," an AI agent that autonomously navigates websites and executes tasks by interpreting screen screenshots to determine precise pixel locations for clicks and keyboard inputs. While the system successfully automates complex workflows like travel planning and safety compliance analysis, it currently faces performance limitations and security vulnerabilities such as prompt injection. This development marks a significant industry shift as competitors race to release similar autonomous agents, signaling a transition toward AI systems that execute entire tasks rather than serving merely as conversational assistants.

Legally bombed: Trump cases dropped

Trump, Rosie Bloor, Adam O'Neill, Dina Musa, Andrew Palmer

Special Counsel Jack Smith moved to dismiss two federal cases against former President Donald Trump, citing presidential immunity and the constitutional policy barring the prosecution of sitting presidents, while Trump simultaneously declared a 25% tariff on Canadian and Mexican imports. Concurrently, a New York court found Trump guilty on all 34 felony counts regarding hush money payments, though sentencing was indefinitely postponed to prevent the conviction of an active president. Legal analysts note that while these federal cases can be refiled depending on future political control, Trump and his allies view the dismissals as a definitive victory against politically motivated prosecutions.

How AI is revolutionising science

Bear, Andy Davis, Alex Zarankov, Emily Shuckborough, Adi Rakhoum, Andy Cooper, Alok Jha

Generative AI is revolutionizing scientific discovery by accelerating drug development, exemplified by Insilico Medicine's rapid creation of an IPF treatment, while simultaneously advancing climate modeling and animal communication research through sophisticated pattern recognition. Autonomous robotic labs and tools like AlphaFold are further democratizing innovation by processing vast datasets and conducting thousands of experiments, though this surge in capability has triggered significant concerns regarding research integrity and AI-generated fraud. As investment flows into these multidisciplinary fields, the industry faces critical challenges in establishing robust data standards and regulations to ensure these tools achieve the projected "golden age" of discovery without compromising scientific validity.

What Trump’s policies could mean for the Fed

Trump, Rob Kaplan, Alison Nathan

In a November 11, 2024 discussion, former Dallas Fed President Rob Kaplan and Goldman Sachs' Alison Nathan evaluate the Federal Reserve's successful "soft landing" execution while highlighting significant purchasing power erosion for the lower 65 million workers. Kaplan predicts distinct regulatory shifts under the new administration and warns that structural factors like labor contraction and potential deregulation will complicate future inflation and growth trajectories. Consequently, he advises markets to treat the anticipated December rate cut as a conditional "game-time decision" pending clearer data on labor dynamics and the administration's fiscal strategy.

Ground assault: Israeli forces invade Lebanon

Rosie Bloor, Greg Karlstrom, Noah Snyder, Andrew Palmer

Israel has initiated a targeted ground incursion into southern Lebanon to dismantle Hezbollah infrastructure and facilitate the return of displaced residents, a move supported by the Biden administration only under strict conditions of limited scope. Simultaneously, Japan's political landscape has shifted as Ishiba Shigeru secured the prime ministership through an unprecedented nine-way contest, signaling a pivot toward conservative liberal policies despite facing an unstable coalition. While these geopolitical maneuvers reshape regional stability, separate corporate commentary highlights the detrimental impact of mandatory "sandwich lunches" on workplace efficiency and employee well-being.

Sudan: inside the world’s worst war

Since April 2024, Rapid Support Forces have besieged Al-Fasher, systematically destroying critical infrastructure and targeting non-Arab Zagawa civilians in acts that experts allege constitute war crimes and genocide. Satellite imagery and human rights investigations confirm the devastation of residential areas, the looting of South Hospital, and the intentional severing of power and water supplies, driving over 100,000 people to flee into conditions of severe malnutrition. With the United Nations declaring famine in nearby camps and international bodies investigating potential atrocities, the crisis threatens to destabilize the broader region while global attention remains insufficient to halt the escalating humanitarian catastrophe.

Is the big tech trade over?

Ben Snider, Chris Hussey

The S&P 500 recently declined 5% following a year-end rally, driven by reassessments of mega-cap valuations amid shifting inflation data and skepticism over AI profitability. Goldman Sachs analysts characterize this correction as a market rotation where capital is flowing from large-cap tech stocks to small-cap equities, a divergence that benefits smaller companies anticipating interest rate cuts. Despite the drop, the firm maintains a positive outlook, citing robust earnings and historical precedents where similar market dips often result in higher valuations within three months.

Regions with OCaml's Local Types | OCaml Unboxed

Jane Street researchers introduce a `regional` variant within the OCaml compiler's `locals` feature that treats function parameters as variables permitted to escape their enclosing region exactly once. This mechanism enforces strict type constraints on return values while enabling critical stack allocation optimizations in recursive tail calls and loop iterations. By defining regions around function and loop bodies, the system prevents $O(N)$ space complexity for local data without compromising the language's type safety guarantees.

The Top 100 GenAI Consumer Apps

Olivia Moore, Steph Smith

Consumer teams will biannually publish a "Generative AI 100" list by screening tens of thousands of data points from SimilarWeb and Sensor Tower to identify the top 50 web and mobile companies. Recent analysis reveals a stark divergence between platforms, with minimal overlap between winners, while mobile trends highlight specific gains in EdTech, avatar generation, and companion apps that rival major social networks in user intensity. Looking forward, the industry is shifting toward a globalized talent base and voice-first interfaces, with Discord community growth serving as a leading indicator for companies like Midjourney and ElevenLabs.