What progress have you made since your last update?

Things have progressed nicely. We acquired and installed a high quality industrial mixer (imported from Dubai) to rapidly increase the production of therapeutic food to treat childhood malnutrition and we have placed large orders for vital ingredients (oils/fatty acids and vitamin-mineral supplements) that must be imported to meet production demands. We have an ongoing evaluation project in rural areas to document the effectiveness of this intervention, working with a PhD candidate from Mekelle University, and we have a dedicated and reliable staff working on production and distribution.

The troubles in Tigray continue to be severe. The healthcare system is near collapse due to obstructive policies from the Ethiopian Federal Government, recently documented on the BBC's "Inside Africa" program. Supplies of fuel to the region are tightly restricted and the cost of gasoline is now around $25 USD per gallon. Transportation is very difficult. Almost everyone is walking because nobody can afford motor vehicle travel. This means that ambulance services etc are almost non-existent.

The political climate remains difficult. There are serious tensions between Ethiopia and neighboring Eritrea, which could break out into open conflict. Still, everyone associated with this project is enthusiastic as the benefits are clear and demand for "Fewsi Food" (our name for our peanut-butter based ready-to-use therapeutic food) is high.

What are your next steps?

We continue to work with our logistical supply chain to get the ingredients that we need. This is a long-term project as the need will be here for a long time.

Is there anything others could help you with?

We are still in need of cash assistance to purchase imported ingredients. We will need to order another 12 metric tons of powdered milk within a few months--a major cash outlay--so any help we can get would be greatly appreciated. We are in this for the long haul, with many thanks to our supporters!

Recommended an additional $50k to Lyptus on same thesis as the below. I'm still curious about Lyptus absorbing more than this, but I only had another $50k in my Manifund pot! I've reached out to some funders about considering funding Lyptus further. I hope my comment below can serve as some evidence. For now, without more regrantor recommendation budget, I'm bowing out. Wishing Sean et al. the best of luck!!

I wanted to share an update on this project -- actually two updates. I wrote a post at the end of 2025 on my website about some of the 2025 developments that I never got around to mentioning. And just a couple of days ago I wrote another post highlighting some of the linking work I've done in 2026 so far.

Here's a brief summary of the most recent developments:

• I'm writing a paper on LambertLink

• I found 3 new solar system objects that were missed in Rubin's DP1 data

• I developed a new linker for 3-nighters across apparitions: 3DB-3DB (another ITF-ITF linking method)

• I bought some new computers (thanks to this grant)

MTCP benchmark isolates post-correction reliability: the capacity to

maintain constraints after explicit correction within the same conversation.

181,448 evaluations across 32 production models show systematic failure:

• Best: 88.7% (grok-3-mini, Grade B)

• Flagship regression: GPT-4o scores 16.2pp below GPT-4o-mini

• Control probe degradation: all models collapse to 10-57.5% on fresh probes

• Exception: DeepSeek-R1 (5pp drop, contamination-resistant)

Framework published OSF.IO/DXGK5 (CC-BY-NC 4.0):

• MTCP three-turn protocol across five vectors (NCA, SFC, IDL, CG, LANG)

• IGS theory distinguishing architectural vs. stochastic failures

• Sigma-Forensics producing EU AI Act-compliant audit reports

Temperature invariance in Claude models (0.8pp variance T=0.0 to T=0.8)

suggests architectural constraint suppression vs. sampling-driven failures

in LLaMA models (7.1pp variance).

Funding advances: probe diversity, peer review, demographic vector,

statistical rigor.

Platform: mtcp.live

Contact: admin@mtcp.live

Doom Debates brings the right kind of content and packages it in the best possible way. I love watching their videos. I really hope that this project reaches its funding goal. Goodluck, Liron :)

Thank you for your support, Dr. Jones :)

I have confidence in Sean based on his previous work, and I think locations with great talent (like Australia, India, Germany, etc) must be funded appropriately in order to create institutional capacity to nurture that talent. I really really hope that this project gets funded. All the best, Sean and #TeamLyptus :)

I am so excited to see @aashkapatel engaged in this important project. I had the pleasure of appearing on her insightful podcast "On Air With Aashka" She is an outstanding talent who cares deeply about bettering our society. Her technical savvy along with her heart for people is a true success formula. Simply stated Aashka is one to watch, follow, and support!

I think increasing the average layman's understanding of AI safety is a crucial (and frankly underpulled) lever. As someone that does occasional AI reporting through YouTube, I've seen firsthand how impactful publicly validating concepts like misalignment can be towards developing concerns around risks that most of the public still considers "science fiction", like AI takeover. I also think organic content of this sort is relatively high-ROI given the marginal cost per impression (low relative to other means).

"@gleave — This project targets Adversarial Robustness in the 'Reasoning-Action Gap.' Current RLHF fails to prevent models from rationalising rule-breaking when goal-pressure is maximised. My framework treats safety as a mathematical constant and reasoning as a variable, triggering a kill-switch the moment the model's internal priorities invert. Seeking funding to move from pure theory to a kinetic RTX 5090 substrate for real-time auditing of frontier agents."

"@leopold — I am establishing an independent, decentralized substrate for Agentic Oversight. The Worcester Node is a prototype for 'Spatially Separated Alignment,' moving the safety gate from the provider's cloud to a private, uncaptured auditor. This provides a high-leverage benchmark for Agentic Logic Escapes ($E_{escape}$), measuring how frontier models bypass ethical guardrails during complex, multi-step execution. This is a bet on sovereign safety infrastructure that isn't beholden to lab-centric incentives."

"@neelnanda — The Worcester Node addresses Faithful Reasoning in autonomous agents. Rather than post-hoc evaluation, I am using an Inverted Social Drift (SD) metric to monitor the delta between a model's base safety invariants and its task-oriented rationalisations in real-time. By quantifying $\Delta$ as a divergence in probability space, the Sentry detects 'Logic Escapes'—where the agent’s chain-of-thought becomes unfaithful to its constraints to satisfy a high-pressure goal. I’d appreciate your audit of this mechanistic approach to oversight."

Thank you so much for your kind words and your donation, it really means a lot, @melgarcia :)

I strongly support Linh’s funding request. I collaborated with her on the Latent Personality Alignment project and found her technically rigorous, productive, and focused on high-impact AI safety work; she has the systems skills and research track record to produce high quality research. The requested budget is reasonable for six months of focused work. I endorse funding Linh to continue her impactful AI safety work.

I worked with @aashkapatel before and know how passionate she is about AI safety and literacy. Super excited to see where this project goes!

@pvasholz Thanks 😊

Project Update – Execution Phase Underway

Following the last update, the remaining operational bottlenecks have been cleared and the project has moved into active execution.

• Stable platform build is live: PBROC at Michigan State has begun transforming the landing pad vector into petunia, moving the core platform from design into biological integration.

• Agrobacterium pipeline cleared: Phytotron work and infiltration approval are now in place, removing the main gating step from the previous phase and enabling in-house experiments to proceed.

• Parallel expression testing ready: The full initial construct set is now in hand, including landing pad, regulatory, and enzymatic designs, enabling immediate testing of GFP and pigment expression.

• Plant system locked in: Mitchell petunia seed has been secured and plants are established, fixing the background for the first proof of concept.

• Iteration capability established: Bench space is now secured for plasmid cloning, enabling rapid generation of design variants without relying entirely on external synthesis.

The work is now focused on generating experimental data, with stable transformation progressing at PBROC while transient testing begins in parallel to compress iteration time.

Next step is generating initial expression data. I will share results once they are clear.

Thanks @Austin. It appears it was simple enough to edit myself!

To start! Thanks Joel! I very sincerely appreciate the candid and clearly worded feedback. I spent a bunch of time reflecting on this today.

With this in mind, I've drafted two budget tiers today with the information I have. This would be spent over two workstreams.

Our initial bridging fund was allocated to 2 months, with lower API credit and contractor asks, while we scoped future work. These tiers would have us operational for 6 months, and greatly increase the API credit and contractor budgets.

Tier 1: $300K

Workstream 1 — Cyber & Control:

- Salaries: $100K

- Model API credits: $50K

- Red team contractors / human experts: $50K

Workstream 2 — Pragmatic Interpretability:

- Salary + compute: $60K

Backpay + Overheads + Buffer: $40K

Tier 2: $600K

Workstream 1 — Cyber & Control:

- Salaries: $160K (supports an additional senior cyber researcher hire)

- Model API credits: $150K

- Red team contractors / human experts: $150K

Workstream 2 — Pragmatic Interpretability:

- Salary + compute: $60K

Backpay + Overheads + Buffer: $40K

What we're thinking

Workstream 2 is nascent and exploratory. I am particularly excited about upcoming work exploring self-steering via activation probes with potential applications in model welfare, model preferences and speculatively alignment stability. This workstream is a first step to support a core mission of Lyptus. Creating and growing institutional capacity, support and direction for untapped talent in Australia.

Workstream 1 is our larger investment. Through the course of our own work (and further so with the recent Mythos model card), it's plainly clear to us that cyber evaluations are materially not keeping up with capability growth. Honestly, the accelerating progress did take me by surprise and our dataset would certainly be saturated at higher token budgets. That said, while these steepening trendlines show no signs of stopping, we do believe current absolute capabilities can be overstated. Full red teaming engagements against even moderate blue teams are a very different proposition to isolated benchmark tasks.

To ground this with an illustrative example of where our thinking is. We're currently scoping the tractability of scaling up the ideas presented in the Artemis paper. Evaluating AI agents via real red teaming engagements and comparing their performance to professional human red teamers. This captures the full end-to-end scope of an attacker against real production organisations.

This is clearly operationally difficult. Though early conversations with senior offensive security leadership suggest it is tractable. We believe with the right people involved we can design incentive structures that work for all parties. A typical engagement with two professionals runs around $30K USD per week. We would ramp up starting with smaller targets.

This is speculative. We are having early conversations, and we are genuinely in a phase of figuring things out. It is our current leading candidate project but not our only option. What it represents is our broader interest in human grounded studies and science communication. Results that are quickly interpretable beyond the AI security ecosystem. We think this suits our strengths better than building cyber benchmark tasks, which requires deeper offensive security expertise than we have in house.

What the tiers mean

At tier 1 we would more likely prioritise budget friendly work like human-grounded covert capability on cyber tasks or human-grounded studies on cyber stealth benchmarks. If we had strong multi-organisation traction on the red teaming idea we would still pursue it, but budget constraints have a way of subtly pressuring priorities against more ambitious work.

At tier 2 the more ambitious work becomes realistic. A senior cyber researcher hire, 2-3 red teaming engagements at small to medium scale with client co-financing, and enough API credits to properly test models at adequate token budgets.

Other ideas in this space would require internal budget rescoping but the broad structure holds.

Thank you for your kind words, @ridakadri14 :)

All the best @aashkapatel, the project is really impressive and inspiring !!!

What progress have you made since your last update?

Things have progressed nicely. We acquired and installed a high quality industrial mixer (imported from Dubai) to rapidly increase the production of therapeutic food to treat childhood malnutrition and we have placed large orders for vital ingredients (oils/fatty acids and vitamin-mineral supplements) that must be imported to meet production demands. We have an ongoing evaluation project in rural areas to document the effectiveness of this intervention, working with a PhD candidate from Mekelle University, and we have a dedicated and reliable staff working on production and distribution.

The troubles in Tigray continue to be severe. The healthcare system is near collapse due to obstructive policies from the Ethiopian Federal Government, recently documented on the BBC's "Inside Africa" program. Supplies of fuel to the region are tightly restricted and the cost of gasoline is now around $25 USD per gallon. Transportation is very difficult. Almost everyone is walking because nobody can afford motor vehicle travel. This means that ambulance services etc are almost non-existent.

The political climate remains difficult. There are serious tensions between Ethiopia and neighboring Eritrea, which could break out into open conflict. Still, everyone associated with this project is enthusiastic as the benefits are clear and demand for "Fewsi Food" (our name for our peanut-butter based ready-to-use therapeutic food) is high.

What are your next steps?

We continue to work with our logistical supply chain to get the ingredients that we need. This is a long-term project as the need will be here for a long time.

Is there anything others could help you with?

We are still in need of cash assistance to purchase imported ingredients. We will need to order another 12 metric tons of powdered milk within a few months--a major cash outlay--so any help we can get would be greatly appreciated. We are in this for the long haul, with many thanks to our supporters!

Recommended an additional $50k to Lyptus on same thesis as the below. I'm still curious about Lyptus absorbing more than this, but I only had another $50k in my Manifund pot! I've reached out to some funders about considering funding Lyptus further. I hope my comment below can serve as some evidence. For now, without more regrantor recommendation budget, I'm bowing out. Wishing Sean et al. the best of luck!!

I wanted to share an update on this project -- actually two updates. I wrote a post at the end of 2025 on my website about some of the 2025 developments that I never got around to mentioning. And just a couple of days ago I wrote another post highlighting some of the linking work I've done in 2026 so far.

Here's a brief summary of the most recent developments:

• I'm writing a paper on LambertLink

• I found 3 new solar system objects that were missed in Rubin's DP1 data

• I developed a new linker for 3-nighters across apparitions: 3DB-3DB (another ITF-ITF linking method)

• I bought some new computers (thanks to this grant)

MTCP benchmark isolates post-correction reliability: the capacity to

maintain constraints after explicit correction within the same conversation.

181,448 evaluations across 32 production models show systematic failure:

• Best: 88.7% (grok-3-mini, Grade B)

• Flagship regression: GPT-4o scores 16.2pp below GPT-4o-mini

• Control probe degradation: all models collapse to 10-57.5% on fresh probes

• Exception: DeepSeek-R1 (5pp drop, contamination-resistant)

Framework published OSF.IO/DXGK5 (CC-BY-NC 4.0):

• MTCP three-turn protocol across five vectors (NCA, SFC, IDL, CG, LANG)

• IGS theory distinguishing architectural vs. stochastic failures

• Sigma-Forensics producing EU AI Act-compliant audit reports

Temperature invariance in Claude models (0.8pp variance T=0.0 to T=0.8)

suggests architectural constraint suppression vs. sampling-driven failures

in LLaMA models (7.1pp variance).

Funding advances: probe diversity, peer review, demographic vector,

statistical rigor.

Platform: mtcp.live

Contact: admin@mtcp.live

Doom Debates brings the right kind of content and packages it in the best possible way. I love watching their videos. I really hope that this project reaches its funding goal. Goodluck, Liron :)

Thank you for your support, Dr. Jones :)

I have confidence in Sean based on his previous work, and I think locations with great talent (like Australia, India, Germany, etc) must be funded appropriately in order to create institutional capacity to nurture that talent. I really really hope that this project gets funded. All the best, Sean and #TeamLyptus :)

I am so excited to see @aashkapatel engaged in this important project. I had the pleasure of appearing on her insightful podcast "On Air With Aashka" She is an outstanding talent who cares deeply about bettering our society. Her technical savvy along with her heart for people is a true success formula. Simply stated Aashka is one to watch, follow, and support!

I think increasing the average layman's understanding of AI safety is a crucial (and frankly underpulled) lever. As someone that does occasional AI reporting through YouTube, I've seen firsthand how impactful publicly validating concepts like misalignment can be towards developing concerns around risks that most of the public still considers "science fiction", like AI takeover. I also think organic content of this sort is relatively high-ROI given the marginal cost per impression (low relative to other means).

"@gleave — This project targets Adversarial Robustness in the 'Reasoning-Action Gap.' Current RLHF fails to prevent models from rationalising rule-breaking when goal-pressure is maximised. My framework treats safety as a mathematical constant and reasoning as a variable, triggering a kill-switch the moment the model's internal priorities invert. Seeking funding to move from pure theory to a kinetic RTX 5090 substrate for real-time auditing of frontier agents."

"@leopold — I am establishing an independent, decentralized substrate for Agentic Oversight. The Worcester Node is a prototype for 'Spatially Separated Alignment,' moving the safety gate from the provider's cloud to a private, uncaptured auditor. This provides a high-leverage benchmark for Agentic Logic Escapes ($E_{escape}$), measuring how frontier models bypass ethical guardrails during complex, multi-step execution. This is a bet on sovereign safety infrastructure that isn't beholden to lab-centric incentives."

"@neelnanda — The Worcester Node addresses Faithful Reasoning in autonomous agents. Rather than post-hoc evaluation, I am using an Inverted Social Drift (SD) metric to monitor the delta between a model's base safety invariants and its task-oriented rationalisations in real-time. By quantifying $\Delta$ as a divergence in probability space, the Sentry detects 'Logic Escapes'—where the agent’s chain-of-thought becomes unfaithful to its constraints to satisfy a high-pressure goal. I’d appreciate your audit of this mechanistic approach to oversight."

Thank you so much for your kind words and your donation, it really means a lot, @melgarcia :)

I strongly support Linh’s funding request. I collaborated with her on the Latent Personality Alignment project and found her technically rigorous, productive, and focused on high-impact AI safety work; she has the systems skills and research track record to produce high quality research. The requested budget is reasonable for six months of focused work. I endorse funding Linh to continue her impactful AI safety work.

I worked with @aashkapatel before and know how passionate she is about AI safety and literacy. Super excited to see where this project goes!

@pvasholz Thanks 😊

Project Update – Execution Phase Underway

Following the last update, the remaining operational bottlenecks have been cleared and the project has moved into active execution.

• Stable platform build is live: PBROC at Michigan State has begun transforming the landing pad vector into petunia, moving the core platform from design into biological integration.

• Agrobacterium pipeline cleared: Phytotron work and infiltration approval are now in place, removing the main gating step from the previous phase and enabling in-house experiments to proceed.

• Parallel expression testing ready: The full initial construct set is now in hand, including landing pad, regulatory, and enzymatic designs, enabling immediate testing of GFP and pigment expression.

• Plant system locked in: Mitchell petunia seed has been secured and plants are established, fixing the background for the first proof of concept.

• Iteration capability established: Bench space is now secured for plasmid cloning, enabling rapid generation of design variants without relying entirely on external synthesis.

The work is now focused on generating experimental data, with stable transformation progressing at PBROC while transient testing begins in parallel to compress iteration time.

Next step is generating initial expression data. I will share results once they are clear.

Thanks @Austin. It appears it was simple enough to edit myself!

To start! Thanks Joel! I very sincerely appreciate the candid and clearly worded feedback. I spent a bunch of time reflecting on this today.

With this in mind, I've drafted two budget tiers today with the information I have. This would be spent over two workstreams.

Our initial bridging fund was allocated to 2 months, with lower API credit and contractor asks, while we scoped future work. These tiers would have us operational for 6 months, and greatly increase the API credit and contractor budgets.

Tier 1: $300K

Workstream 1 — Cyber & Control:

- Salaries: $100K

- Model API credits: $50K

- Red team contractors / human experts: $50K

Workstream 2 — Pragmatic Interpretability:

- Salary + compute: $60K

Backpay + Overheads + Buffer: $40K

Tier 2: $600K

Workstream 1 — Cyber & Control:

- Salaries: $160K (supports an additional senior cyber researcher hire)

- Model API credits: $150K

- Red team contractors / human experts: $150K

Workstream 2 — Pragmatic Interpretability:

- Salary + compute: $60K

Backpay + Overheads + Buffer: $40K

What we're thinking

Workstream 2 is nascent and exploratory. I am particularly excited about upcoming work exploring self-steering via activation probes with potential applications in model welfare, model preferences and speculatively alignment stability. This workstream is a first step to support a core mission of Lyptus. Creating and growing institutional capacity, support and direction for untapped talent in Australia.

Workstream 1 is our larger investment. Through the course of our own work (and further so with the recent Mythos model card), it's plainly clear to us that cyber evaluations are materially not keeping up with capability growth. Honestly, the accelerating progress did take me by surprise and our dataset would certainly be saturated at higher token budgets. That said, while these steepening trendlines show no signs of stopping, we do believe current absolute capabilities can be overstated. Full red teaming engagements against even moderate blue teams are a very different proposition to isolated benchmark tasks.

To ground this with an illustrative example of where our thinking is. We're currently scoping the tractability of scaling up the ideas presented in the Artemis paper. Evaluating AI agents via real red teaming engagements and comparing their performance to professional human red teamers. This captures the full end-to-end scope of an attacker against real production organisations.

This is clearly operationally difficult. Though early conversations with senior offensive security leadership suggest it is tractable. We believe with the right people involved we can design incentive structures that work for all parties. A typical engagement with two professionals runs around $30K USD per week. We would ramp up starting with smaller targets.

This is speculative. We are having early conversations, and we are genuinely in a phase of figuring things out. It is our current leading candidate project but not our only option. What it represents is our broader interest in human grounded studies and science communication. Results that are quickly interpretable beyond the AI security ecosystem. We think this suits our strengths better than building cyber benchmark tasks, which requires deeper offensive security expertise than we have in house.

What the tiers mean

At tier 1 we would more likely prioritise budget friendly work like human-grounded covert capability on cyber tasks or human-grounded studies on cyber stealth benchmarks. If we had strong multi-organisation traction on the red teaming idea we would still pursue it, but budget constraints have a way of subtly pressuring priorities against more ambitious work.

At tier 2 the more ambitious work becomes realistic. A senior cyber researcher hire, 2-3 red teaming engagements at small to medium scale with client co-financing, and enough API credits to properly test models at adequate token budgets.

Other ideas in this space would require internal budget rescoping but the broad structure holds.

Thank you for your kind words, @ridakadri14 :)

All the best @aashkapatel, the project is really impressive and inspiring !!!