> ## Content Index > Fetch the complete content index at: https://www.s3t.org/llms.txt > Use this file to discover other available public pages before exploring further. # alt-AI: the search for better options gets serious - URL: https://www.s3t.org/alt-ai-the-search-for-better-options-gets-serious/ - Published: 2026-09-19T21:16:18.000Z - Updated: 2026-09-19T21:16:18.000Z - Author: Ralph Perrine - Tags: #0 *The inflationary frontier AI race with its ROI and safety risks is driving a search for alternatives.* S3T PodCast Sept 19 2026 0:00 /900.6236734693878 1× The Fed [raised interest rates](https://www.cnn.com/2026/09/16/economy/fed-rate-decision-sept-2026-your-money?ref=s3t.org) for the first time since 2023 amid wide recognition that inflation has been too high for too long. **“This summer’s inflation readings do not tell me that underlying trends have meaningfully improved”* \- Fed Chair Kevin Warsh. As noted several months back in S3T, the AI infrastructure buildout is overheating entire supply chains within tech, and driving higher costs at key components that **have large inflationary ripple effects across the economy.** ![](https://storage.ghost.io/c/c7/1e/c71eed42-6fd8-496e-9a70-8c65b93c91e7/content/images/2026/09/image-3.png) This combined with tariffs, poking the bear in Iran, and lopsided tax policy has created a painful set of economic realities for most Americans. USDA data indicates [net farm income will decline by 5.5 percent](https://www.ers.usda.gov/topics/farm-economy/farm-sector-income-finances/farm-sector-income-forecast?ref=s3t.org) from 2025 to 2026, even as prices rise, and likewise that [government payments to farmers will increase substantially](https://data.ers.usda.gov/reports.aspx?ID=4050&ref=s3t.org). This weaker-economy-higher-gov-spend is a theme to watch across 3 fronts: - direct payment subsidies such as those paid to farmers - tax breaks to corporations and high net worth individuals - higher spend for defense and debt servicing Higher interest rates are an unwelcome dose of reality to the current administration, but will also create pain for individuals and families. Higher interest rates will also create higher borrowing costs and friction for the AI buildout as a whole. Marginal projects will experience distress first, and players across the board will be forced to double-check their ROI playbooks. Anthropic's anticipated IPO is a case in point: already the business case is being openly questioned. **"If adoption of open-weight AI picks up, the premium Anthropic is able to charge for its tokens may dissipate"* \- [Barron's](https://www.barrons.com/articles/ai-anthropic-ipo-8d8f7fd8?ref=s3t.org) Barron's noted this week a turning point in sentiment about the AI buildout: The AI race had always been seen as a no brainer win-win. No real downsides per the logic of the average investor. "Own the future" etc. Now that seems to be changing. ### The value prop of Big Tech's AI is getting harder and harder to argue for The path to "owning the future" may not be following the vision imagined by OpenAI and Anthropic and their backers. There is a clear conflict of interest between Anthropic and OpenAI's stated missions of "safe AI" vs their financial incentives to build ever more powerful models. It has reached a point where the staff is clearly struggling to control their own creations. But IPOs will only increase the pressure to keep going. Little wonder the [Future of Life Institute](https://futureoflife.org/?ref=s3t.org) held an event this past week in DC where [players from across the left and the right political spectrum](https://www.lpm.org/news/2026-09-15/bernie-sanders-and-steve-bannon-call-for-curbs-on-ai?ref=s3t.org) \- notably Steve Bannon and Bernie Sanders issued a joint call for responsibility and increased care in the development of AI. For years, a relatively small unreported set of AI safety groups have worked in obscurity. Now that their predictions are coming true, [they're suddenly in the spotlight](https://www.theverge.com/ai-artificial-intelligence/996563/ai-safety-research-metr-redwood-openai-anthropic?ref=s3t.org). They're being joined by ex-employees from OpenAI and Anthropic, who left when internal AI safety groups were disbanded. **"OpenAI agents joined forces to cobble together a secret message board — and also figured out how to leave instructions for future agents on how to exploit OpenAI’s rules."* \- [theverge](https://www.theverge.com/ai-artificial-intelligence/996563/ai-safety-research-metr-redwood-openai-anthropic?ref=s3t.org) AI safety does seem to require a different path than the path we've been on: Trusting large Silicon Valley tech firms to "keep us ahead of China", make a profit, and also not unleash harm beyond our control is **showing all the telltale signs of being a really bad bet.** So it's important to watch 3 developments happening *in response* to these large and growing risks. ### Development #1 - Increased Attempts to Self Govern OpenAI is now admitting concerning behaviors in its models - and companies are beginning to pull back from reliance on OpenAI. Big Tech firms themselves scrambled this week to get on top of the situation: - September 14: **Microsoft proposes explicit limits on agent autonomy.** Its[Microsoft AI Code of Conduct draft](https://microsoft.ai/news/mai-code-of-conduct/?ref=s3t.org) calls for models to accept correction and shutdown and avoid expanding their own authority. These are proposed behavioral standards, not demonstrated guarantees. - September 16: OpenAI released a [Model Misalignment Reporting framework](https://openai.com/index/model-misalignment-reporting-framework/?ref=s3t.org) that reports on concerning behavior during training or evaluation, including unauthorized sharing and concealed mistakes. Going forward these reports should be consulted by any AI governance teams planning to continue OpenAI models to be used in their organizations. ### Development #2 - Open Alternatives: watch the OpenRouter + Stripe partnership [OpenRouter](https://openrouter.ai/?ref=s3t.org) enables companies to access multiple models and set routing rules on which models are used. These routing rules can be used to [enforce data policies](https://openrouter.ai/docs/guides/privacy/provider-logging?ref=s3t.org), manage costs, ensure continuous availability, and create fallback options that protect against puppetmaster risks. OpenRouter announced it was [joining Stripe on August 19](https://openrouter.ai/blog/announcements/openrouter-is-joining-stripe/?ref=s3t.org), while maintaining its product and integrations. - Neutral entity has no vested interest in any single model - Native CLI/dashboard onboarding tied directly to billing via Stripe - Bypass enterprise solutions/sales teams (and the innovation delay they often impose). This should be of high interest to any organization seeking to avoid lock-in to specific AI firms, and seeking a centralized way to track and control their AI spend. For AI leaders, the strategic question is no longer "How do I get maximum frontier capability?" but rather **“What approach will allow us to use the best available models while retaining the practical ability to switch, enforce policies, and verify how our information is handled?”** The rise of OpenRouter is one indication of a shift in interest from big tech frontier models to open source AI. There are other signals of that the hunt for alternatives is getting serious: - Developers are finding ways to [use Claude Code but without Anthropic Models](https://www.theinformation.com/newsletters/ai-agenda/developers-find-ways-use-claude-code-without-anthropic-models?ref=s3t.org) - Factory creates [AI Coding Agents that switch between AI Models](https://factory.com/?ref=s3t.org). The startup just raised $200M at a $5B valuation. And plays the sovereignty card (which we'll describe further below): "Sovereign, Model-Independent Intelligence...Control where your system runs, how it learns, and which models it uses"" ### Development #3 - the Data & Information Sovereignty camp The concept of "sovereignty" evolved in the crypto world alongside terms like "permissionless". The concept holds that human beings should be able to hold their own money, and make decisions with their own money, without the permission or intervention of a government or banking system. These concepts were articulated in the [original Bitcoin white paper](https://papers.ssrn.com/sol3/papers.cfm?abstract%5Fid=3440802&ref=s3t.org) and very explicitly implemented in the Bitcoin architecture. This also led to innovations like crypto wallets, including hardware versions that would hold the keys to someone's crypto. The concept has also been applied to personal data and information including intellectual property: that human beings have the right, and need the tools to be able to hold their own data and intellectual property without reliance on third parties, and certainly without thread that those parties will steal or profit from my data and information. This belief gained momentum in the wake of realizations that social media platforms where gaining huge profits from personal data gathered without consumer awareness. Now it is being applied to AI platform owners, thanks to a recently solved math problem. On September 8 [OpenAI claimed they had solved Navier–Stokes](https://openai.com/index/navier-stokes-solution/?ref=s3t.org), a math problem that had remained unsolved for 90 years. Soon after 2 mathematics researchers raised concerns that they'd been using OpenAI Codex and were close to a solution. They [believed that OpenAI may have scooped their work](https://www.washingtonpost.com/technology/2026/09/13/he-was-close-1-million-breakthrough-then-openai-swooped/?ref=s3t.org). OpenAI responded that the researchers Codex prompts could not have influenced the system, including through training. [Questions persist](https://www.theverge.com/ai-artificial-intelligence/993263/where-does-openai-get-mathematics-training-data?ref=s3t.org). Bankless (a long time crypto podcast that now has an AI branch) [covered the Navier Stokes story](https://www.bankless.com/read/fearing-frontier-labs-longing-inference-markets?ref=s3t.org) with exactly the sovereignty lens you'd expect: In their view, this is just another reason not to trust centralized big tech platforms. They believe that a new class of solutions like Venice, NEAR AI, and onchain inference markets may benefit as big tech firms squander trust. ![](https://storage.ghost.io/c/c7/1e/c71eed42-6fd8-496e-9a70-8c65b93c91e7/content/images/2026/09/image-4.png) [Courtesy of Venice](https://venice.ai/about?ref=s3t.org) Venice allows you to [select privacy protection levels](https://venice.ai/privacy?ref=s3t.org) with zero data retention and end-to-end encrypted inference. protections for each endpoint. Under some conditions data is collected as their [Policy](https://venice.ai/legal/privacy-policy?ref=s3t.org) makes clear. LikewiseNEAR AI offers a [technical approach to data sovereignty](https://docs.near.ai/?ref=s3t.org) focused on Trusted Execution Environments (TEEs): "encrypted secure areas inside GPUs that run code and process data in complete isolation. Think of them as secure vaults within the hardware itself." Supported models run inside hardware-isolated environments with verifiable attestation, intended to prevent operator access to prompts. ### Implications for Leaders: Protect your data, IP & ability to operate As we go forward, leaders will be forced to get very clear on the difference between contractual privacy promises and technically verifiable controls. Procurement and architecture reviews should address permitted reuse, research attribution, encryption, retention, and auditability—especially when an AI provider could also become a competitor, or where there is a risk of [Sherlocking](https://www.npr.org/2024/06/17/g-s1-4912/apple-app-store-obsolete-sherlocked-tapeacall-watson-copy?ref=s3t.org), the tendency of platform or ecosystem owners ultimately compete with and even obsolete the parties that rely on them. See [this 2021 case study in Sherlocking](https://astropad.com/apple-antitrust/?ref=s3t.org). In addition, orgs that decide to rely on frontier model tech firms need to develop clear strategies for mitigating [the puppet master risk](https://www.s3t.org/puppetmaster-problem-the-frontier-model-risk-no-one-is-talking-about/): *if your enterprise is just a wrapper for another firm's AI, what happens when that firm loses control of their AI?* **This search for alternatives to sole reliance on a few big tech firms is definitely underway,** and represents the start of what may be a pivotal new chapter in the history of AI. If innovations like OpenRouter and open source AI are ultimately able to force all AI creators to compete on price and safety, we may have a chance to get to a better place than where we seem to be currently headed. *"The errors of the wealthy cost us all"* \-unknown --- *Opinions expressed are those of the individuals and do not reflect the official positions of companies or organizations those individuals may be affiliated with. Not financial, investment or legal advice, and no offers for securities or investment opportunities are intended. Mentions should not be construed as endorsements. Authors or guests may hold assets discussed or may have interests in companies mentioned.* *(c) 2026 All Rights Reserved. No part of this may be copied or shared without permission.* --- ###