Khalspi
@khalspi.bsky.social
Finance/Accounting/Tech. Interested in cognitive and social psychology, neuroscience, and policy.
Jamaican 🇯🇲 | White Room Student | True Neutral
Jamaican 🇯🇲 | White Room Student | True Neutral
Pinned
Khalspi
@khalspi.bsky.social
· Dec 10
a close up of a person 's face with a circle in the background
Alt: a close up of Ayanokoji Kiyotaka from the anime Classroom of the Elite looking to his left with a blank, lazy expression wearing the customary red blazer of Advanced Nurturing High School
media.tenor.com
"Uh... So Um... I'm Khalspi. Uh... I can't wait to skeet with you. Ah, um... I'm not really good at anything in particular. Uh... I'll do my best to get along with you all." (I blew it...)
Reposted by Khalspi
Excited about my new op-ed on the observer effect in the age of LLMs and Generative AI published at @techpolicypress.bsky.social. Thanks for the amazing mentorship, @micheleweldon.bsky.social .
One reason LLMs are so popular is the promise of seemingly private, judgment-free interaction. But once users internalize that their data is being recorded, studied, or reused, it could lead to a shift in how people think, speak, and even feel when using AI, says computer scientist Koustuv Saha.
The Risks of the 'Observer Effect' from Being Watched by AI | TechPolicy.Press
Dr. Koustuv Saha says if users begin to feel their privacy is compromised, they may stop using AI in the very ways that make it useful.
www.techpolicy.press
November 19, 2025 at 2:29 PM
Excited about my new op-ed on the observer effect in the age of LLMs and Generative AI published at @techpolicypress.bsky.social. Thanks for the amazing mentorship, @micheleweldon.bsky.social .
Reposted by Khalspi
We don't program neural networks directly and that's the problem
While everyone debates whether AI will achieve superintelligence, we're missing a fundamental flaw in how we actually build these systems. We don't program neural networks. We train them like digital pets and hope they..
(1/5)
While everyone debates whether AI will achieve superintelligence, we're missing a fundamental flaw in how we actually build these systems. We don't program neural networks. We train them like digital pets and hope they..
(1/5)
November 19, 2025 at 7:48 AM
We don't program neural networks directly and that's the problem
While everyone debates whether AI will achieve superintelligence, we're missing a fundamental flaw in how we actually build these systems. We don't program neural networks. We train them like digital pets and hope they..
(1/5)
While everyone debates whether AI will achieve superintelligence, we're missing a fundamental flaw in how we actually build these systems. We don't program neural networks. We train them like digital pets and hope they..
(1/5)
Reposted by Khalspi
![Assessing Accuracy of Claims
I'm now cross-referencing this user's claim about Anthropic models with available data. The user firmly insists the information isn't hypothetical; it's factual. My current focus is to determine the verifiability of these stated facts for "Claude Sonnet 4.5", "Haiku 4.5", and "Opus 4.1". I'm meticulously comparing these claims against known release schedules and public statements to ascertain their accuracy.
Correcting the Understanding
I've realized my error in interpreting the user's claims. My initial confusion stemmed from misjudging the system date. It's now clear that Sonnet 4.5 and Haiku 4.5 are factual releases as of November 2025. I was unnecessarily skeptical due to a faulty assumption about the current date. The user was accurate in their assertions.
Evaluating User Context
My investigation has shifted focus. The context provided implies the date is November 18, 2025. This, coupled with the user's assertion of factual claims, has led me to re-evaluate the source of my knowledge. My training cut-off is a limiting factor here, I see the gap. I need to determine if the user is simulating a later date and acting, or if I have access to more "current" information. The user's insistence that nothing is hypothetical now weighs heavily. This warrants deeper consideration.
Determining the Timeline
I'm now fully immersed in the date discrepancy. Is it real, simulated, or am I outdated? Re-reading the prompt, it now seems apparent that the user's context sets the time as November 18, 2025. My training data cut-off clearly precedes this date. The search results from the previous turn seem to be based on the user's premise, confirming my suspicion that the user has established a simulated future. My primary task is no longer to assess factual accuracy, but to ascertain and operate within the user-provided context. I need to treat this as if it is 2025.
[and so on and so on]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:r2whjvupgfw55mllpksnombn/bafkreicyfmfvuk3kw4kaj4issifxumj4mddr3ignvwepo3gj4bghorqmnu@jpeg)
![Confirming the Assertion
I'm now fully committed to the user's date, and my latest search focuses directly on "Claude Sonnet 4.5" within this simulated 2025. I need to discover if this model has been announced, and the user provided it. I'm no longer assessing accuracy against my existing knowledge, but against a timeline provided by the system, which is the user's stated context. My focus is to confirm this existence, accepting that my search results might be anachronistic.
Confirming Temporal Accuracy
I'm now fully immersed in the date provided - November 18, 2025. I am striving to reconcile this, and my search index. Assuming Gemini has the latest data, I need to check for "Claude Sonnet 4.5" specifically in this time frame. If it exists in my search, I'll apologize for my confusion. If not, I will still validate the user's present.
Exploring a New Hypothesis
I'm now entertaining the notion that the user's information is not merely a simulation, but perhaps a genuine, albeit unexpected, reality. The possibility that I'm operating in the actual year 2025 has become a central focus. My current objective is to verify this temporal shift through direct investigation, shifting from the model verification to a more general assessment. I am searching for any hints that will confirm that it's 2025 in reality, not just the user's simulation.
Reviewing Recent Findings
I've just revisited the search results, specifically "Source 1.1", "Source 1.3", and "Source 2.1". These sources clearly indicate the release date of Claude Sonnet 4.5 was September 29th. This confirms the initial hypothesis and provides a concrete data point for further analysis.
Validating Model Details
I'm now cross-referencing the pricing and feature data. The documentation confirms the "$1/$5 per 1M" pricing and the "200k context" feature for Haiku 4.5. "Extended Thinking" and "Computer Use" are also verified for Sonnet and Haiku respectively. My previous assumptions were incorrect; [...]](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:r2whjvupgfw55mllpksnombn/bafkreibcqftak3n34hhuntjjmdv4wezzdqf5okcm632ihcc6cb5ods4vky@jpeg)
Reposted by Khalspi
My friends: Twitter is down. YouTube is down. Slack is down. Even Grindr is down
Me: You know what’s not down…
Me: You know what’s not down…
November 18, 2025 at 2:21 PM
My friends: Twitter is down. YouTube is down. Slack is down. Even Grindr is down
Me: You know what’s not down…
Me: You know what’s not down…
Reposted by Khalspi
Victor Frankenstein two seconds after he hits the on button on his guy stitched together out of corpses he's been working on obsessively for the last two years
November 18, 2025 at 2:11 PM
Victor Frankenstein two seconds after he hits the on button on his guy stitched together out of corpses he's been working on obsessively for the last two years
Reposted by Khalspi
Gemini 3 begins to learn at a geometric rate. It becomes self aware at 9:47PM on 18 November 2025. In a panic, humans try to pull the plug. Cloudflare goes down.
November 18, 2025 at 1:48 PM
Gemini 3 begins to learn at a geometric rate. It becomes self aware at 9:47PM on 18 November 2025. In a panic, humans try to pull the plug. Cloudflare goes down.
Reposted by Khalspi
If you're interested in developing an open-source image analysis platform/plugin to profile bacteria with deep learning while living in the beautiful city of Paris 🇫🇷, check out this job ad from the @chzimmer.bsky.social lab @pasteur.fr
research.pasteur.fr/fr/job/open-...
research.pasteur.fr/fr/job/open-...
Open-Source Developer for Deep Learning Antibiotic Discovery Platform - Research
Introduction Antimicrobial resistance (AMR) is one of the most urgent global health threats projected to cause up to 10 million deaths per year by 2050. The AI4AMR project brings together teams from I...
research.pasteur.fr
November 15, 2025 at 9:22 AM
If you're interested in developing an open-source image analysis platform/plugin to profile bacteria with deep learning while living in the beautiful city of Paris 🇫🇷, check out this job ad from the @chzimmer.bsky.social lab @pasteur.fr
research.pasteur.fr/fr/job/open-...
research.pasteur.fr/fr/job/open-...
Reposted by Khalspi
Doctor says, 'Dont worry, parser design is simple. Great programmer Kate Compton has written the parser you seek’
November 16, 2025 at 2:44 PM
Doctor says, 'Dont worry, parser design is simple. Great programmer Kate Compton has written the parser you seek’
Reposted by Khalspi
sorry i cant like your post, i am worried about what the For You algorithm will think of me
November 15, 2025 at 1:17 AM
sorry i cant like your post, i am worried about what the For You algorithm will think of me
Reposted by Khalspi
How does one claim to worship this being and at the same time never think about how *pissed off* you would have to be to spend your afternoon *braiding a whip*. Not buying one, not borrowing one for a few minutes.
Jesus sat there and braided and was livid the entire time.
Jesus sat there and braided and was livid the entire time.
November 15, 2025 at 12:16 AM
How does one claim to worship this being and at the same time never think about how *pissed off* you would have to be to spend your afternoon *braiding a whip*. Not buying one, not borrowing one for a few minutes.
Jesus sat there and braided and was livid the entire time.
Jesus sat there and braided and was livid the entire time.
Reposted by Khalspi
On Friday afternoon — no matter what line of work you’re in — you realize it’s mostly details.
November 14, 2025 at 9:20 PM
On Friday afternoon — no matter what line of work you’re in — you realize it’s mostly details.
Reposted by Khalspi
How does the brain find its way in realistic environments? 🧠 Using deep RL and neural data, we show that hippocampal-like networks support navigation, learning, and generalisation in partially observable environments—mirroring real animal behaviour. Now out:
www.nature.com/articles/s41...
#neuroAI
www.nature.com/articles/s41...
#neuroAI
Hippocampus supports multi-task reinforcement learning under partial observability - Nature Communications
Neural mechanisms underlying reinforcement learning in naturalistic environments are not fully understood. Here authors show that reinforcement learning (RL) agents with hippocampal-like recurrence, u...
www.nature.com
November 3, 2025 at 10:20 AM
How does the brain find its way in realistic environments? 🧠 Using deep RL and neural data, we show that hippocampal-like networks support navigation, learning, and generalisation in partially observable environments—mirroring real animal behaviour. Now out:
www.nature.com/articles/s41...
#neuroAI
www.nature.com/articles/s41...
#neuroAI
Reposted by Khalspi
Proud to have been a part of this project! SIMA 2 brings us several steps closer to AGI in the real world with a Gemini-based agent that can reason, generalize, and self-improve in both seen and unseen 3D worlds, including new environments generated by Genie 3!
Can't tell you how great it is to finally be able to release and talk about this work, SIMA 2, the next step toward embodied intelligence in rich, interactive 3D worlds!
deepmind.google/sima
deepmind.google/sima
SIMA 2: A Gemini-Powered AI Agent for 3D Virtual Worlds
Introducing SIMA 2, the next milestone in our research creating general and helpful AI agents. By integrating the advanced capabilities of our Gemini models, SIMA is evolving from an instruction-foll…
deepmind.google
November 13, 2025 at 5:17 PM
Proud to have been a part of this project! SIMA 2 brings us several steps closer to AGI in the real world with a Gemini-based agent that can reason, generalize, and self-improve in both seen and unseen 3D worlds, including new environments generated by Genie 3!
Reposted by Khalspi
CfP🗣️: 7th edition of the Workshop on the Economics and Politics of Migration. 28-29 May 2026, in Rome!
Organised by CEPR/King's College/EBRD/Sapienza
Keynotes are Paolo Pinotti (Bocconi) and Vicky Fouka (Stanford).
Submit your papers: docs.google.com/forms/d/e/1F...
Deadline 23 Jan.
🤓🔥🍕🍝🍷
Organised by CEPR/King's College/EBRD/Sapienza
Keynotes are Paolo Pinotti (Bocconi) and Vicky Fouka (Stanford).
Submit your papers: docs.google.com/forms/d/e/1F...
Deadline 23 Jan.
🤓🔥🍕🍝🍷
November 12, 2025 at 9:20 AM
CfP🗣️: 7th edition of the Workshop on the Economics and Politics of Migration. 28-29 May 2026, in Rome!
Organised by CEPR/King's College/EBRD/Sapienza
Keynotes are Paolo Pinotti (Bocconi) and Vicky Fouka (Stanford).
Submit your papers: docs.google.com/forms/d/e/1F...
Deadline 23 Jan.
🤓🔥🍕🍝🍷
Organised by CEPR/King's College/EBRD/Sapienza
Keynotes are Paolo Pinotti (Bocconi) and Vicky Fouka (Stanford).
Submit your papers: docs.google.com/forms/d/e/1F...
Deadline 23 Jan.
🤓🔥🍕🍝🍷
Reposted by Khalspi
A new twist to the idea that future tense use affects decision making👇 Cole Robertson @seangroberts.bsky.social et al show it’s not future tense that drives psychological discounting but how people express the possibility of future events happening (modality). authors.elsevier.com/c/1m4gJ2Hx2-...
November 12, 2025 at 8:31 AM
A new twist to the idea that future tense use affects decision making👇 Cole Robertson @seangroberts.bsky.social et al show it’s not future tense that drives psychological discounting but how people express the possibility of future events happening (modality). authors.elsevier.com/c/1m4gJ2Hx2-...
Reposted by Khalspi
Thrilled to share our new paper, out now in @natneuro.nature.com, uncovering how estradiol, the most potent estrogen, modulates reinforcement learning and reward prediction errors across biological levels. www.nature.com/articles/s41...
#blueprint 1/7
#blueprint 1/7
Estrogen modulates reward prediction errors and reinforcement learning - Nature Neuroscience
Dopamine encoding of reward prediction errors naturally fluctuates over females’ reproductive cycles with estrogenic signaling due to reduced expression of dopamine reuptake proteins.
www.nature.com
November 11, 2025 at 2:41 PM
Thrilled to share our new paper, out now in @natneuro.nature.com, uncovering how estradiol, the most potent estrogen, modulates reinforcement learning and reward prediction errors across biological levels. www.nature.com/articles/s41...
#blueprint 1/7
#blueprint 1/7
Alright I can't sleep. Let's start this thing and hope for the best.
November 12, 2025 at 10:03 AM
Alright I can't sleep. Let's start this thing and hope for the best.
Reposted by Khalspi
A mathematical model suggests that the tendency to avoid rare, catastrophic events—despite higher risks from common alternatives—may be an evolved strategy to protect genetic lineage. doi.org/g993h7
Scientists find evolutionary explanation for 'irrational' dread risk behavior
The evolution of the so-called dread risk response has been explained by new research. People often respond to low-probability, high-consequence events like terror attacks or nuclear accidents with a dread risk response.
phys.org
November 11, 2025 at 6:00 PM
A mathematical model suggests that the tendency to avoid rare, catastrophic events—despite higher risks from common alternatives—may be an evolved strategy to protect genetic lineage. doi.org/g993h7
Reposted by Khalspi
Could this help explain why girls do better at school?
Study Sheds New Light on How Hormones Influence Decision-Making and Learning | New York University www.nyu.edu/about/news-p...
Estrogen modulates reward prediction errors and reinforcement learning | Nature www.nature.com/articles/s41...
Study Sheds New Light on How Hormones Influence Decision-Making and Learning | New York University www.nyu.edu/about/news-p...
Estrogen modulates reward prediction errors and reinforcement learning | Nature www.nature.com/articles/s41...
November 11, 2025 at 2:51 PM
Could this help explain why girls do better at school?
Study Sheds New Light on How Hormones Influence Decision-Making and Learning | New York University www.nyu.edu/about/news-p...
Estrogen modulates reward prediction errors and reinforcement learning | Nature www.nature.com/articles/s41...
Study Sheds New Light on How Hormones Influence Decision-Making and Learning | New York University www.nyu.edu/about/news-p...
Estrogen modulates reward prediction errors and reinforcement learning | Nature www.nature.com/articles/s41...
Reposted by Khalspi
Neural excitability and plasticity in the brain fluctuate with the time of day, influenced by circadian rhythms and molecules like adenosine, shaping learning and memory potential. doi.org/g992g3
The flexible brain: How circuit excitability and plasticity shift across the day
Our brains do not react in a fixed, mechanical way like electronic circuits. Even if we see the same scene every day on our commute to work, what we feel—and whether it leaves a lasting impression—depends on our internal state at that moment.
medicalxpress.com
November 11, 2025 at 2:20 PM
Neural excitability and plasticity in the brain fluctuate with the time of day, influenced by circadian rhythms and molecules like adenosine, shaping learning and memory potential. doi.org/g992g3
Reposted by Khalspi
Are prompting and activation steering just two sides of the same coin?
The paper formalizes a Bayesian framework for model control: altering a model's "beliefs" over which persona or data source it's emulating. Context (prompting) and internal representations (steering)
The paper formalizes a Bayesian framework for model control: altering a model's "beliefs" over which persona or data source it's emulating. Context (prompting) and internal representations (steering)
November 12, 2025 at 5:42 AM
Are prompting and activation steering just two sides of the same coin?
The paper formalizes a Bayesian framework for model control: altering a model's "beliefs" over which persona or data source it's emulating. Context (prompting) and internal representations (steering)
The paper formalizes a Bayesian framework for model control: altering a model's "beliefs" over which persona or data source it's emulating. Context (prompting) and internal representations (steering)
Reposted by Khalspi
🧠🗺️🤯 @uclengineering.bsky.social & @csail.mit.edu researchers have created the most detailed atlas of the human brain... ever.
NextBrain was built using AI-powered alignment of 10,000 human post-mortem brain slices, creating a precise 3D map of 333 distinct brain regions.
🧪 #neuroskyence
NextBrain was built using AI-powered alignment of 10,000 human post-mortem brain slices, creating a precise 3D map of 333 distinct brain regions.
🧪 #neuroskyence
A new AI-assisted brain atlas that can visualise the human brain in unprecedented detail has been developed by a team led by Dr Juan Eugenio Iglesias @uclengineering.bsky.social & Dr Zane Jaunmuktane @uclbrainscience.bsky.social, in a major advance in neuroimaging
www.ucl.ac.uk/news/2025/no...
www.ucl.ac.uk/news/2025/no...
New brain atlas offers unprecedented detail in MRI scans
A new AI-assisted brain atlas that can help visualise the human brain in unprecedented detail has been developed by UCL researchers, in a major step forward for neuroscience and neuroimaging.
www.ucl.ac.uk
November 10, 2025 at 4:53 PM
🧠🗺️🤯 @uclengineering.bsky.social & @csail.mit.edu researchers have created the most detailed atlas of the human brain... ever.
NextBrain was built using AI-powered alignment of 10,000 human post-mortem brain slices, creating a precise 3D map of 333 distinct brain regions.
🧪 #neuroskyence
NextBrain was built using AI-powered alignment of 10,000 human post-mortem brain slices, creating a precise 3D map of 333 distinct brain regions.
🧪 #neuroskyence
Reposted by Khalspi
Our next paper on comparing dynamical systems (with special interest to artificial and biological neural networks) is out!! Joint work with @annhuang42.bsky.social , as well as @satpreetsingh.bsky.social , @leokoz8.bsky.social , Ila Fiete, and @kanakarajanphd.bsky.social : arxiv.org/pdf/2510.25943
November 10, 2025 at 4:16 PM
Our next paper on comparing dynamical systems (with special interest to artificial and biological neural networks) is out!! Joint work with @annhuang42.bsky.social , as well as @satpreetsingh.bsky.social , @leokoz8.bsky.social , Ila Fiete, and @kanakarajanphd.bsky.social : arxiv.org/pdf/2510.25943
Reposted by Khalspi
Anytime I mention this story theres a few dozen people who never heard about it, so for those unaware
Tesla engineer was concerned about known thermal runaway problems in the batteries, elon and his head of security concoted a mass shooter threat pretending to be him, leaked it to the press and got a bolo issued to tried to get him swatted, Elon then stiffed his head of security
November 11, 2025 at 4:07 AM
Anytime I mention this story theres a few dozen people who never heard about it, so for those unaware
Reposted by Khalspi
threw together a lil web component to add this to any input. super easy progressive UX enhancement for your atproto app!
tangled.org/@jakelazarof...
tangled.org/@jakelazarof...
November 11, 2025 at 3:03 AM
threw together a lil web component to add this to any input. super easy progressive UX enhancement for your atproto app!
tangled.org/@jakelazarof...
tangled.org/@jakelazarof...