Mar 10, 2025

Preparing for Accelerated AGI Timelines

Sofia Mendez, Grace Gong, Mahi Shah

This project examines the prospect of near-term AGI from multiple angles—careers, finances, and logistical readiness. Drawing on various discussions from LessWrong, it highlights how entrepreneurs and those who develop AI-complementary skills may thrive under accelerated timelines, while traditional, incremental career-building could falter. Financial preparedness focuses on striking a balance between stable investments (like retirement accounts) and riskier, AI-exposed opportunities, with an emphasis on retaining adaptability amid volatile market conditions. Logistical considerations—housing decisions, health, and strong social networks—are shown to buffer against unexpected disruptions if entire industries or locations are suddenly reshaped by AI. Together, these insights form a practical roadmap for individuals seeking to navigate the uncertainties of an era when AGI might rapidly transform both labor markets and daily life.

Reviewer's Comments

Reviewer's Comments

Arrow
Arrow
Arrow
Arrow
Arrow

Hi Sofia, Grace, and Mahi, I really enjoyed reading your project and appreciate the practical relevance of it, that’s the greatest strength of your project. If you’re going forward with it, I see this fitting either as a blogpost or a peer-reviewed article. A few recommendations to strengthen your work:

(1) Add citations in text, it supports your argument and offers a point of reference to the reader. It also clearly distinguishes between what might be your perspective and that of the others/what is backed up by evidence.

(2) While there was a good mix of resources, there is an over-reliance on a single source, which is LessWrong. To strengthen your argument and credibility, I’d try to strike a balance between multiple sources - you can even look historically at academic articles or case studies about how technological advancements have impacted society, and provide that as further insight into the possible effects of AI.

(3) Expand by providing a more detailed, step-by-strep process as to how you did the review in the section “b. Implementation”. Imagine if you were the reader and trying to replicate this, what would you need to write to be clear cut to you?

(4) For greater impact on the reader, add some thought-provoking quotes from the sources you mentioned in section “3. Findings”.

(5) Really go in depth into the implications of your findings, as that will add a lot of practical support to your idea.

Strengths:

- Proactive Approach: The topic is highly relevant, addressing societal and economic implications of accelerated AGI timelines.

- Identification of Challenges: The paper effectively identifies key challenges, such as labor market impacts and investment changes.

- Thematic Organization: Clear categorization into career, financial, and logistical preparedness is a logical structure.

Areas for Improvement:

- Add in-text citations to distinguish between original ideas and existing research.

- Include a detailed methodology section explaining how the literature review was conducted.

- Propose actionable recommendations or novel insights based on the literature reviewed.

- Suggest solutions or mitigation strategies for the challenges identified.

- Analyze limitations and potential negative consequences of the reviewed literature.

- Clearly outline the methodology for reproducibility.

- Improve in-text citations and ensure all sources are properly referenced.

- Expand the discussion to include your own perspective, recommendations, and future research directions.

Suggestions for Future Work:

- Conduct a more systematic literature review, including peer-reviewed research.

- Explore cross-disciplinary perspectives (e.g., ethics, policy) to broaden the scope and impact.

Clearly written and easy to follow, and an interesting perspective.

The main feedback I have is that it seems strange to me to frame this as a problem to be approached on an individual level, rather than on a societal level. It also seems unlikely that such an individual solution will help even the individual for more than a quite limited time. I would have been more excited about a version that focused on what we could to to strengthen resilience on society level (or perhaps if this individual preparation was targeted for a specific group which would be particularly important to keep well prepared so that they can keep up important safety related work)

Cite this work

@misc {

title={

Preparing for Accelerated AGI Timelines

},

author={

Sofia Mendez, Grace Gong, Mahi Shah

},

date={

3/10/25

},

organization={Apart Research},

note={Research submission to the research sprint hosted by Apart.},

howpublished={https://apartresearch.com}

}

Recent Projects

View All

View All

Feb 2, 2026

Markov Chain Lock Watermarking: Provably Secure Authentication for LLM Outputs

We present Markov Chain Lock (MCL) watermarking, a cryptographically secure framework for authenticating LLM outputs. MCL constrains token generation to follow a secret Markov chain over SHA-256 vocabulary partitions. Using doubly stochastic transition matrices, we prove four theoretical guarantees: (1) exponentially decaying false positive rates via Hoeffding bounds, (2) graceful degradation under adversarial modification with closed-form expected scores, (3) information-theoretic security without key access, and (4) bounded quality loss via KL divergence. Experiments on 173 Wikipedia prompts using Llama-3.2-3B demonstrate that the optimal 7-state soft cycle configuration achieves 100\% detection, 0\% FPR, and perplexity 4.20. Robustness testing confirms detection above 96\% even with 30\% word replacement. The framework enables $O(n)$ model-free detection, addressing EU AI Act Article 50 requirements. Code available at \url{https://github.com/ChenghengLi/MCLW}

Read More

Feb 2, 2026

Prototyping an Embedded Off-Switch for AI Compute

This project prototypes an embedded off-switch for AI accelerators. The security block requires periodic cryptographic authorization to operate: the chip generates a nonce, an external authority signs it, and the chip verifies the signature before granting time-limited permission. Without valid authorization, outputs are gated to zero. The design was implemented in HardCaml and validated in simulation.

Read More

Feb 2, 2026

Fingerprinting All AI Cluster I/O Without Mutually Trusted Processors

We design and simulate a "border patrol" device for generating cryptographic evidence of data traffic entering and leaving an AI cluster, while eliminating the specific analog and steganographic side-channels that post-hoc verification can not close. The device eliminates the need for any mutually trusted logic, while still meeting the security needs of the prover and verifier.

Read More

Feb 2, 2026

Markov Chain Lock Watermarking: Provably Secure Authentication for LLM Outputs

We present Markov Chain Lock (MCL) watermarking, a cryptographically secure framework for authenticating LLM outputs. MCL constrains token generation to follow a secret Markov chain over SHA-256 vocabulary partitions. Using doubly stochastic transition matrices, we prove four theoretical guarantees: (1) exponentially decaying false positive rates via Hoeffding bounds, (2) graceful degradation under adversarial modification with closed-form expected scores, (3) information-theoretic security without key access, and (4) bounded quality loss via KL divergence. Experiments on 173 Wikipedia prompts using Llama-3.2-3B demonstrate that the optimal 7-state soft cycle configuration achieves 100\% detection, 0\% FPR, and perplexity 4.20. Robustness testing confirms detection above 96\% even with 30\% word replacement. The framework enables $O(n)$ model-free detection, addressing EU AI Act Article 50 requirements. Code available at \url{https://github.com/ChenghengLi/MCLW}

Read More

Feb 2, 2026

Prototyping an Embedded Off-Switch for AI Compute

This project prototypes an embedded off-switch for AI accelerators. The security block requires periodic cryptographic authorization to operate: the chip generates a nonce, an external authority signs it, and the chip verifies the signature before granting time-limited permission. Without valid authorization, outputs are gated to zero. The design was implemented in HardCaml and validated in simulation.

Read More

Feb 2, 2026

Markov Chain Lock Watermarking: Provably Secure Authentication for LLM Outputs

We present Markov Chain Lock (MCL) watermarking, a cryptographically secure framework for authenticating LLM outputs. MCL constrains token generation to follow a secret Markov chain over SHA-256 vocabulary partitions. Using doubly stochastic transition matrices, we prove four theoretical guarantees: (1) exponentially decaying false positive rates via Hoeffding bounds, (2) graceful degradation under adversarial modification with closed-form expected scores, (3) information-theoretic security without key access, and (4) bounded quality loss via KL divergence. Experiments on 173 Wikipedia prompts using Llama-3.2-3B demonstrate that the optimal 7-state soft cycle configuration achieves 100\% detection, 0\% FPR, and perplexity 4.20. Robustness testing confirms detection above 96\% even with 30\% word replacement. The framework enables $O(n)$ model-free detection, addressing EU AI Act Article 50 requirements. Code available at \url{https://github.com/ChenghengLi/MCLW}

Read More

Feb 2, 2026

Prototyping an Embedded Off-Switch for AI Compute

This project prototypes an embedded off-switch for AI accelerators. The security block requires periodic cryptographic authorization to operate: the chip generates a nonce, an external authority signs it, and the chip verifies the signature before granting time-limited permission. Without valid authorization, outputs are gated to zero. The design was implemented in HardCaml and validated in simulation.

Read More

Feb 2, 2026

Markov Chain Lock Watermarking: Provably Secure Authentication for LLM Outputs

We present Markov Chain Lock (MCL) watermarking, a cryptographically secure framework for authenticating LLM outputs. MCL constrains token generation to follow a secret Markov chain over SHA-256 vocabulary partitions. Using doubly stochastic transition matrices, we prove four theoretical guarantees: (1) exponentially decaying false positive rates via Hoeffding bounds, (2) graceful degradation under adversarial modification with closed-form expected scores, (3) information-theoretic security without key access, and (4) bounded quality loss via KL divergence. Experiments on 173 Wikipedia prompts using Llama-3.2-3B demonstrate that the optimal 7-state soft cycle configuration achieves 100\% detection, 0\% FPR, and perplexity 4.20. Robustness testing confirms detection above 96\% even with 30\% word replacement. The framework enables $O(n)$ model-free detection, addressing EU AI Act Article 50 requirements. Code available at \url{https://github.com/ChenghengLi/MCLW}

Read More

Feb 2, 2026

Prototyping an Embedded Off-Switch for AI Compute

This project prototypes an embedded off-switch for AI accelerators. The security block requires periodic cryptographic authorization to operate: the chip generates a nonce, an external authority signs it, and the chip verifies the signature before granting time-limited permission. Without valid authorization, outputs are gated to zero. The design was implemented in HardCaml and validated in simulation.

Read More

This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.
This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.