In this issue:
- USAID Under Fire
- AST2VEC: Teaching Machines to Read Code Like Humans
- Big Tech Critic Takes Key FCC Role
- Nockchain Technical Update
USAID Under Fire
Historically unprecedented developments have thrust the United States Agency for International Development (USAID) into a remarkable public controversy. The agency is now facing major allegations about its operations, spending, and true mission.
Elon Musk and the DOGE team gained access to USAID's systems after officials who tried to block access were placed on leave. On X, Musk labeled USAID a "criminal organization" and exposed extraordinary sums of wasteful spending and propaganda funding. His team's audit led to widespread operational disruptions, including staff being ordered to work remotely and the agency's website being taken down.
Former State Department official Mike Benz has shown that USAID operates as a "geopolitical manipulation arm" with a $50 billion annual budget exceeding both the CIA and State Department combined. He alleges the agency's main purpose is to fund political movements, media outlets, and censorship campaigns—at home and abroad—under the guise of humanitarian aid.
AST2VEC: Teaching Machines to Read Code Like Humans
Since our ultimate mission is to deliver simple tools for networked tribes to achieve independence, we’re paying close attention to how AI is likely to intersect with the NockApp framework.
AST2VEC is a neural network (2021) that translates Python code into 256-dimensional vectors and back again, trained on nearly half a million student programs. Unlike previous approaches that simply counted code patterns or measured edit distances, AST2VEC learns to understand program structure while preserving enough information to reconstruct the original code.
The system shows impressive capabilities on beginner Python programs, perfectly reconstructing most programs under 36 nodes (about 79% of student submissions). It processes code quickly—encoding takes about 0.9ms per 10 nodes, while decoding takes 2.8ms per 10 nodes.
The key innovation is treating code like a language with grammar rules, allowing the neural network to learn meaningful representations while ensuring all decoded programs remain syntactically valid.
Big Tech Critic Takes Key FCC Role
Adam Candeub, a prominent critic of Big Tech and advocate for social media regulation, has been appointed as general counsel of the Federal Communications Commission. FCC Chairman Brendan Carr praised Candeub as a "fearless" fighter against "Big Tech censorship" with experience in both telecom litigation and academia.
Candeub rose to prominence through his work challenging social media companies' content moderation practices. He has advocated for regulating major platforms as common carriers—requiring political viewpoint neutrality—and for narrowing Section 230 protections. During the Trump administration, he served as acting head of the National Telecommunications and Information Administration, where he pushed for similar reforms.
This appointment comes alongside other key Trump nominations targeting tech power, including antitrust critic Gail Slater to lead the DOJ's antitrust division. Candeub's vision calls for expanding antitrust law to address tech companies' influence over "democratic institutions such as free speech" and "the marketplace of ideas."
Nockchain Technical Update
As part of our ongoing internal testing, we recently created a NockApp that queries multiple Nockchain nodes for their current consensus state. This has allowed us to see multiple nodes attain consensus with much faster block times than what we intend for mainnet.
We've also been refining our protocol for propagating unconfirmed transactions to miners. By moving the onus to the wallet to keep the network aware of an unconfirmed transaction, we’ve been able to both limit the number of unconfirmed transactions any node needs to store, and allow for transaction replacement to address “stuck” transactions.
