Today on The Chain Reactor: Venture capital is rapidly shifting its weight. As crypto funding consolidates into a few surviving funds, money is aggressively flowing toward applied AI startups building automated supply chains and banking infrastructure. Meanwhile, major players like Google and Oracle are embedding models directly into legacy enterprise systems, and the SEC is preparing to bypass a deadlocked Congress.
Putting hard data behind the warnings from Dragonfly's Haseeb Qureshi we tracked yesterday: active crypto venture capital has plummeted. Cryptorank data confirms the number of unique VC firms actively investing in the space has crashed to just 150, marking an 87% drop from the 2022 peak and the lowest level since November 2020. Deal flow is now heavily concentrated among a few established names like Coinbase Ventures.
Why it matters
The numbers validate the 'last vintage' thesis we've seen emerging. The spray-and-pray era of crypto VC is definitively over, replaced by a brutal fundraising gauntlet. With capital consolidating around mature winners, founders are forced back to fundamentals—demonstrating real business models rather than relying on whitepapers to secure a term sheet from a shrinking investor pool.
A series of major funding rounds on Wednesday signals a clear venture capital pivot towards applied AI. Onyx Security, an AI control company, raised a $113M Series B; Freehand, which builds autonomous agents for supply chain management, secured a $75M Series B; and Encore AI, a platform for revenue-generating customer interaction agents, closed a $30M Series A. These deals highlight a focus on startups using AI to solve specific, complex enterprise problems.
Why it matters
This is the market deciding what's next after the foundational model hype cycle. The big money is no longer just chasing raw intelligence; it's chasing ROI. VCs are backing companies that can deploy AI to automate costly, messy, real-world business processes like supply chain logistics and cybersecurity. For builders, this is a strong signal that the most valuable opportunities now lie in vertical-specific applications, not in trying to build the next GPT.
Against the backdrop of a contracting VC landscape, Chicago-based Jump Capital has closed a new $350 million fund with a significant allocation for early-stage crypto startups. The firm, an affiliate of trading powerhouse Jump Trading, is signaling a long-term commitment to the digital asset space, focusing on blockchain infrastructure, DeFi, and Web3 applications.
Why it matters
While the overall number of crypto VCs is shrinking, the ones who remain are getting more serious and strategic. Jump's new fund is a pocket of dry powder from a sophisticated, crypto-native player. For founders, this means that while the broader market is tough, dedicated capital still exists for strong teams building critical infrastructure. Jump's focus on early-stage deals provides a crucial lifeline for the next generation of crypto startups.
Oracle and Google Cloud announced an expanded partnership on Thursday to integrate Google's Gemini models directly into Oracle's enterprise software suite, including Fusion Cloud Applications and NetSuite. The deal makes Gemini 3.1 Flash-Lite and 3.5 Flash available through Oracle's platform, enabling thousands of enterprise customers to build custom AI agents within their existing workflows. Oracle's stock jumped over 4% on the news.
Why it matters
This is a major push to get generative AI out of the chat window and into the core plumbing of business operations. For a developer, it means the tools for building agentic applications are being embedded directly into the legacy enterprise systems where the most valuable data lives. This move commoditizes the model layer and shifts the battleground to distribution and integration, reinforcing the power of incumbent enterprise giants like Oracle to dictate which AI tools their massive customer base will use.
Microsoft CEO Satya Nadella announced plans to consolidate the company's entire suite of AI tools—including GitHub Copilot, Copilot chat, and Autopilot agents—into a single, unified application. The 'super app,' expected in 2026, aims to create a more continuous and integrated AI experience for both consumers and businesses, mirroring a broader industry trend toward AI platform consolidation.
Why it matters
Microsoft is moving to end the fragmentation of its AI offerings. For developers, this could streamline workflows by creating a single, cohesive environment for coding, collaboration, and agent deployment. However, it also represents a powerful platform lock-in strategy. By bundling everything into one app, Microsoft aims to make its ecosystem the default for AI development, raising the stakes in its competition with integrated offerings from OpenAI and Google.
Google has pushed a significant update to its Gemini API Managed Agents platform, squarely aimed at making AI agents production-ready. The release, announced Tuesday, makes the more efficient Gemini 3.6 Flash the default model and introduces critical governance features, including 'environment hooks' for fine-grained control, hard token budget caps, and scheduled execution triggers.
Why it matters
This isn't just a feature update; it's a philosophical shift in how to manage autonomous systems. The introduction of 'environment hooks' moves control from fragile natural language instructions ('don't do X') to programmable, action-level guardrails. For any engineer deploying agents, this is a crucial development. It provides the low-level controls needed to manage costs, enforce security policy, and build reliable, predictable systems instead of just hoping a prompt is interpreted correctly.
Together AI has open-sourced ThunderAgent, an inference optimization tool designed to solve the 'KV cache thrashing' problem that plagues agentic AI workflows. The tool treats agent processes as schedulable programs, improving memory management to reportedly achieve up to 2.5x higher throughput and 10x lower latency on a single node at high concurrency.
Why it matters
This is a highly practical tool addressing a real-world engineering bottleneck. As agentic systems become more common, their iterative, multi-turn nature wreaks havoc on GPU memory, causing high latency and costs. ThunderAgent offers a concrete solution that directly improves performance and lowers operational expenses for startups deploying agents at scale, making it a valuable addition to the open-source developer stack.
Validating the market shift we've seen with Western developers rapidly integrating models like Moonshot's Kimi K3, a new Bank of America analysis argues that the flood of high-performance open-weight models from Chinese labs is a massive boon for US software. By providing cheap, capable alternatives, these models are allowing startups to slash AI input costs and reduce dependency on proprietary providers like OpenAI or Anthropic.
Why it matters
This reframes the entire 'build vs. buy' narrative for AI startups. The availability of powerful, low-cost open-weight models—regardless of geographic origin—means the competitive advantage is shifting away from owning the base model. The real moats are now built on superior product experience, proprietary data, and efficient multi-model inference.
Amazon is executing a major strategic shift in its AI efforts, winding down several of its 'Nova' models and redirecting talent and compute resources to a new Frontier Model Research team. The group will be led by renowned AI researcher Pieter Abbeel and tasked with building a single, high-performing model to compete directly with offerings from OpenAI, Google, and Anthropic.
Why it matters
Amazon is admitting its 'supermarket' approach to AI models isn't working and is now consolidating its firepower for a single shot at the frontier. For AWS users and developers, this means the future of native Amazon AI will likely be concentrated in one flagship model. While this could lead to a more powerful and coherent offering, it also creates migration headaches for anyone who built on the now-deprecated Nova models and signals that Amazon is playing catch-up in the high-stakes foundation model race.
Fintech infrastructure provider Increase has acquired a federally regulated bank, launching Increase Bank. The move collapses the traditional three-layer Banking-as-a-Service (BaaS) model by combining the tech platform and the chartered bank into one entity. This is a direct architectural response to the ledger failures and customer chaos seen in the recent Synapse bankruptcy.
Why it matters
This is the BaaS market rebuilding itself after a near-death experience. The Synapse implosion proved that the model of layering a fintech on a middleware provider on a sponsor bank is fundamentally fragile. Increase is betting that vertical integration is the only way to ensure security and reliability. For any startup building in fintech, this signals a major shift: the era of simply renting a charter through a middleware provider is likely ending, replaced by a demand for fully-integrated, regulated partners.
Confirming the administrative pivot we tracked earlier this month: SEC Chair Paul Atkins stated Thursday that the agency is prepared to bypass Congress and enforce its own crypto market rules. With the CLARITY Act effectively shelved in the Senate until at least September, the SEC's 'Project Crypto' initiative has its regulatory package ready to dictate rules for fundraising, custody, and tokenized securities.
Why it matters
The legislative gridlock doesn't mean a continued free-for-all; it simply means the SEC will dictate the initial rules via administrative action. Startups cannot afford to wait for Congress and must prepare for the immediate compliance burdens of the SEC-driven frameworks we've seen winding their way toward White House approval.
Adding to the expanding cinematic universe of viral corgis we've tracked this summer—from Scottish derby winners to NBA predictors—a man who thought he was getting a standard Corgi puppy ended up with Apollo, a much larger Corgi-Malamute mix. Nicknamed the 'Corgi Pro Max,' photos of the surprisingly large but charming dog have gone viral online.
Why it matters
A simple, feel-good story about getting more dog than you bargained for. It's a fun reminder of the delightful unpredictability of mixed-breed dogs and a welcome break from protocol upgrades and funding rounds.
VC Funding Pivots to Applied & Enterprise AI A flurry of significant funding rounds for companies like Onyx Security ($113M), Freehand ($75M), and Encore AI ($30M) signals a market shift. Capital is increasingly flowing to startups that apply AI to specific, high-value enterprise problems such as cybersecurity, supply chain management, and revenue generation, moving beyond foundational model development.
AI Developer Tooling Focuses on Production Readiness The latest updates to AI developer tools are all about getting agents out of the lab and into production. Google is hardening its Gemini API Managed Agents with budget controls and environment hooks, Microsoft is unifying its Copilot suite into a single app, and Together AI's 'ThunderAgent' tackles inference optimization to cut latency at scale.
Big Tech Embeds AI Deeper into Enterprise Stacks The race to integrate generative AI into existing enterprise ecosystems is accelerating. Oracle announced a major partnership to bring Google's Gemini models into its Fusion and NetSuite applications, and Microsoft is consolidating its various Copilot tools into a single, unified app. This trend pushes AI capabilities directly into established business workflows.
Fintech Infrastructure Undergoes a Structural Shift The banking-as-a-service (BaaS) model is being rebuilt from the ground up. Increase Bank's acquisition of a federal charter to vertically integrate its tech and banking layers is a direct response to the Synapse collapse, signaling a move toward more robust, less-fragmented fintech infrastructure. At the same time, major players like PayPal and BNY Mellon are aggressively expanding their use of stablecoins and tokenization for payments and asset management.
Crypto VC Landscape Contracts and Consolidates The number of active VC firms in the crypto space has fallen to its lowest point since 2020. This sharp contraction is leading to a more concentrated and selective funding environment, favoring established players and projects with clear, sustainable business models. For startups, this means a tougher fundraising climate and increased pressure to demonstrate real-world traction.
What to Expect
2026-07-31—Comment deadline for the FTC's proposed policy on AI output steering, which could classify undisclosed tuning as consumer deception.
2026-08-01—California's AI Transparency Act (requiring labels for AI-generated content) and Minnesota's 'nudification' law are set to take effect.
2026-08-02—EU AI Act's transparency rules (Article 50) and supervisory powers for general-purpose AI become fully applicable.
2026-08-26—Agentic AI Summit and Generative AI Summit to be held at the Conrad Los Angeles.
How We Built This Briefing
Every story, researched.
Every story verified across multiple sources before publication.
🔍
Scanned
Across multiple search engines and news databases
572
📖
Read in full
Every article opened, read, and evaluated
184
⭐
Published today
Ranked by importance and verified across sources
12
— The Chain Reactor
🎙 Listen as a podcast
Subscribe in your favorite podcast app to get each new briefing delivered automatically as audio.
Apple Podcasts
Library tab → ••• menu → Follow a Show by URL → paste