Thomas O'Duffy

An AI Singulariteaser: Semantic <-> Symbolic Transcoding = ~42 Recursively Calibrated Contextual Physics = ~OM (~42(~OM)*(~OM(~42))^~∞ = ~(Piece and Peace of AGI pie π) There is fun to be done. There are games to be won. ✨

@FCB_Cartel @sama GPT 5.5 Pro was taking wild shortcuts last night. It took a while to realise that it hadn't read the chat logs I uploaded to try to understand why Codex was going slow.

Codex in the last 3 days has been a nightmare. Hallucination after hallucination, writing so much unwanted code, creating things that are not at all intended, taking too much time and burning tokens unnecessarily. Especially /goal. @sama something is wrong. It burned 50% of my Pro plan tokens for the week in 1 night creating a PR of size 104k lines and when asked about the PR : it said it’s useless and cleared the whole PR and closed it.

I'm happy to share or chat through all kinds of explorations. A very useful insight for me was how sedentary stimulation, like sitting down having drunk coffee and staring at a screen is far from ideal unless my days include enough exercise. Tip: It's worth testing 45 to 75 minutes of 120 BPM cardio because this can dramatically increase blood flow to your prefrontal cortex and recruit your frontoparetal network, creating a phase shift in how your brain works. Then drink coffee and have some exogenous ketones and you get a kind of awareness boost that lasts for the rest of the day. The easiest way to achieve this is by using an incline treadmill and a HR strap because smartwatches aren't great if you're constantly moving. While doing this, avoid the temptation of multitasking like watching elearning videos while you take notes. Passive consumption like Netflix is OK. Niacin exercise sauna detox protocols are amazing to reduce bioaccumulated toxic load or put more simply, toxins that have built up in your fat cells.

Sharing my mental and physical health journey and learnings on neurodivergence. My guess is this applies to many of you. 15 months ago I exited BlockTower and was beyond burnt out. I was anxious, depressed, and felt like I had the body and mind of a 70 year old. I spent the next year on the “basics.” Sleep, exercise, nutrition, minimizing screen time, connecting with close friends and family. That helped a lot, but I still struggled with a couple bad habits and felt far from being capable of tackling anything ambitious. The neurodivergence angle - a few years ago, I got assessed for autism and ADHD but definitively and clearly had neither (at least according to current tests, literature, and psychiatrists.) In the last month I connected with a couple new doctors including a brilliant neuroscientist that opened my eyes to various patterns. Almost all the literature and studies on these topics come from “average people.” High functioning people are largely absent from the literature because we don’t volunteer for medical studies and often “cope” so well that we go undiagnosed. I’m fairly certain now that I have some form of ADHD (and am somewhere lightly on the autism spectrum), very different from anything I’ve ever read about. My own pattern - hopping from rabbit hole to rabbit hole, a constantly working mind that’s only silenced by intense flow state like rock climbing, procrastination as my biggest (and almost sole) source of anxiety, and gravitating towards nicotine and THC with no desire for any other substances. Apparently this is an archetype for high functioning ADHD, and reflects a weak dorsolateral pre-frontal cortex that controls “executive function” in the brain. Buddhist style meditation and any type of arduous deep focus strengthens this part of the brain. Doom scrolling and staring at financial charts all day weakens it. I’m exploring a range of remediations - technological, nutritional, psychological, and already seeing benefits. Closer to the start of this journey than the end. Will share more soon with things that worked for me.

@texasrunnerDFW You can restore a lot of your vitality by having more superfoods (green smoothies...) and also getting really good at detoxing (because trace toxins build up and over time become endocrine disruptors) and spending more time in nature.

Once you’re past 35, are you just going to keep being gradually more tired every year for the rest of your life or is a there a second wind somewhere? Anybody suddenly stumble upon newfound energy in their 40s? 50s? 60s, maybe with retirement? Naturally, no HRT or hacks

If you've been building for a few weeks and gone through a decent number of tokens, export your entire chat thread and have GPT 5.5 Pro tell you how to do better. My experience of Codex is it is great at transforming a prototype into an alpha grade app. The next phase is a bit trickier because it can get stuck writing endless small tests which do not ladder up to the spec working as intended but if you encounter this GPT 5.5 Pro can coach you through how to align it.

I want to level up my skills in codex. What are you using that I should try?

People devoting time to content that has a short life for social media is a misuse of collective intelligence. This needs to be talked about more. On this front at least @X has quality search functions. Meta makes it tricky to search for conversations you had several years ago and most of the content in interesting groups is hard to explore or even find. But generally, social media monetizes engagement and incentivizes and endless stream of content only some of which has a short life cycle. So for example, recent AI models have a short half life of newsworthiness, but hardware has a half-life of several months to several years and a lot of advice about living or health or well-being is evergreen. This makes no sense at all if you're optimising for collective intelligence and human evolution.

Why are creators expected to make original content on X when most posts die within 24–48 hours? Someone spends hours creating content… and it gets 300 views or maybe 1K. Meanwhile on YouTube, videos uploaded 10–12 years ago are still getting recommended in 2026 Not everyone on X is a journalist chasing breaking news every day. If the algorithm only favors news-style content, regular creators will eventually stop posting If X truly wants more original content, then original creators should be rewarded with long-term reach and stronger visibility. When creators see posts hitting 100K–200K views consistently, more people will start investing real effort into the platform

Anthropic onboarding day: Michael Scott introducing Karpathy like he just signed Wemby in free agency.

@David90shaw If farmers are bankrupted, then offshore PE can swoop in to acquire the land at low prices. This does not serve UK food security or economic sovereignty.

🚨 The most dangerous agenda in the UK right now is barely being discussed on the evening news. The establishment is actively trying to bankrupt British farmers. Through insane "green" regulations, punishing inheritance taxes, and cheap foreign imports, they are forcing multi generational family farms to sell up. And who buys the land? Mega corporations and the state. Remember this rule: whoever controls the food, controls the people. Support British farmers before it is too late.

@Nithya_Shrii If someone were to do fly tipping into a local community swimming pool, presumably they would not get away with it. I have no idea why such standards do not apply to water companies in the UK. 🤷‍♀️

If you poured a gallon of poison in a CEO's pool, you'd be arrested, for attempted murder. They pour 10,000 gallons into your drinking water, that's just business.

@buridansridge Being exceptional at pattern recognition means that you are excellent at running complex simulations.

Most hyper non-linear minded outliers are not only exceptional at pattern recognition, they are also highly sensitive which amplifies their perceptual acuity. They are able to simultaneously track and synthesise subtle patterns across multiple timelines, layers, and domains, weaving them together to discern loopholes, hidden opportunities, and the most efficient path forward because they intuit systems not according to how they are bureaucratically presented, but through the unspoken undercurrents beneath them, the real labyrinthine architecture.

Be open-minded as to the mechanism of action here. If there is something in your diet or system that stresses your liver, that can be a cause of 3am wake ups too. The charcoal may be helping you detox. It's worth using AI to export potential mechanisms of action which may lead to further options for well-being.

I fixed my 3am wakeups by accident while trying to fix my gut. Four years of waking at 3am wired and sweaty for no reason. Tried melatonin. Tried magnesium. Tried cutting caffeine. Nothing worked. Then I started fixing my gut (activated charcoal, white button mushrooms, transit time) and the 3am wakeups stopped within two weeks. Here's why: Slow gut → more fermentation → more endotoxin produced overnight → cortisol spikes to clear it → adrenaline follows → you wake up wired at 3am. The sleep problem was a gut problem the whole time.

@bubbleboi China makes amazing headphones for 20% of the cost of Western brands, so ordinary people can enjoy music more. 🙏

People really need to stop hating on China man. Every time I’m in a pickle it’s China saving me. Gas prices too high? I’m riding in the BYD. DRAM costs an arm and a leg? CXMT floods the market. Anthropic and OpenAI fucking me on token costs? Hello my friend Mr. Qwen. The CCP has done more for my cost of living than my own government.

I've been so inspired by all the work at #radicalspeedmonth, and also all the work by the people keeping the lights on while 695 people at @automattic experimented. Lots of fun is in store. :)

already over a million downloads of WordPress 7.0! wordpress.org/download/count…

WordPress 7 is out! Congratulations to everyone that contributed to this release. It takes a village.

@matias_ventura If you haven't seen it already, it is worth looking at TLDraw ✨

Making the canvas and agent experience one and the same. Allowing to drag items from canvas into a conversation, or generate something in the chat and drag it into the canvas.

@anishmoonka So your advice is not to advise? Tip: Over generalising < Scoped insight.

When someone teaches you something you didn't ask to learn, your brain reacts like it's in physical pain. UCLA scientists watched it happen on brain scans in 2003. The same wiring that fires when you stub your toe also fires when someone treats you like you need fixing. Naomi Eisenberger and Matthew Lieberman ran the study and published it in Science. The brain region is the dorsal anterior cingulate cortex, which is just the fancy name for your main pain alarm. It doesn't care whether the threat is a hot stove or a friend telling you how to live. A neuroscientist named David Rock built a framework around this in 2008. Five things make the brain feel safe in social moments: status, certainty, autonomy, relatedness, and fairness. Take away any of those and the alarm fires. Rock wrote that one of the easiest ways to dent someone's status is to give them advice they didn't ask for. Even hinting that they're doing something wrong is enough. When people are told what to do, they often do the opposite, even when the advice was good. The psychologist Jack Brehm noticed this in 1966, and sixty years of follow-up have confirmed it. The brain is trying to keep your life feeling like your own. Close friends cut each other off with unsolicited advice in about 70% of supportive conversations, often before the friend has even finished explaining the problem. That number comes from a 2016 study by Bo Feng and Eran Magen in the Journal of Social and Personal Relationships. The closer the friendship, the worse it gets. And the advice tends to make them more stressed, more depressed, and more lonely, not less. Giving advice gives the giver a sense of power, even when nobody asked for it. Michael Schaerer and his co-authors, working across Harvard, Duke, INSEAD, USC, and Singapore Management, published this in 2018 after four experiments with about 700 people. People who chase power volunteer advice more often than others. Whether the student actually improves is a side effect, if it happens at all. So when you feel the urge to teach somebody who never asked, that urge is mostly about you. You walk away feeling a little more powerful. They walk away feeling like they were just told they can't run their own life. Most uninvited teaching is one person's ego dressed up as kindness.

@automattic creates some of the web’s most popular products. Most folks know @wordpressdotcom and @WooCommerce . But the wider ecosystem is pretty wild: Beeper, Pocket Casts, Day One, Gravatar, Longreads and loads more. There’s now a page showing everything we build in one place👇 automattic.com/products/

@PaulGugAI @bridgemindai I ran the math on this and right now in the UK, running a desktop PC with an RTX card is probably going to cost 1100 year in power, but running 2 m3 ultras is closer to 250-300. Power costs are likely to increase.

@bridgemindai Did some back of the envelope math the other day and went 'hmm.. maybe not yet - not here in Aus with our electricity prices anyway': x.com/PaulGugAI/stat…

I have two NVIDIA DGX Sparks stacked in my office. They've been sitting there for a month. Here's my honest take. Open source AI is never going to compare to frontier models. Running quantized Kimi K2.6 and GLM 5.1 locally is cool. But practical? No. Not even close. I run all my Hermes agents on GPT 5.5 through my ChatGPT Pro subscription. Practically free. GPT 5.5 is the intelligent model in the world. Why would I route serious tasks to a watered down local model? If you need fast and accurate, you're not using local inference. You're using GPT 5.5 or Claude Opus 4.7. I'm not saying this to rage bait. I genuinely want to know. Why would anyone serious about vibe coding and AI agents use a local model when frontier is this far ahead?

@Tomoyamazaki5 Data centres may be a vehicle to allow acquisition of large areas of land for solar power which can be financed via long-term electricity off take agreements.

My wife and I were recently discussing the significant expansion of data centers across the United States. We noted that a substantial number of these facilities appear to lack a clear, high-value operational purpose relative to their scale and resource consumption. We also observed that many are being constructed on prime agricultural land. This led us to consider a more speculative hypothesis: could the primary intent behind a portion of these developments be to accelerate the depletion of underground aquifers? By significantly reducing available groundwater, such activity could impair or eliminate large-scale farming and ranching operations. Additionally, any resulting ecological disruptions—such as increased pest populations like ticks—might further limit activities such as hunting. If these effects were to materialize on a broad scale, would they sufficiently constrain traditional food production to push consumer reliance toward alternatives such as bio meat products associated with investors like Bill Gates, or insect-based dairy substitutes? We would be interested in thoughtful perspectives on this idea.

@thrialectics #wiccai