My other option would be to upgrade the graphics card on my desktop, but that does not really solve the integration elements of the agent that I prefer within the Apple ecosystem. I know that the Hermes Agent development is primarily driven within that environment and that explains the integration being more useful. I decided to pay for some tokens from the Portal ecosystem Nous Research runs to get a feel for how the Claude Opus 4.8 model performs. To be fair it really is awesome. Going back to using the Qwen models locally using Ollama really does feel like a step backward. After you get a feel for how the cutting edge models perform it is easy to see that the next generation of models are going to be pretty epic.

It’s interesting to see how the model generations are translating to how much VRAM they need to run effectively locally using something like Ollama. A lot of the graphics cards that were sold for gaming cap out at 8 gigabytes of VRAM which means when the models shifted to being larger in terms of parameters a generation of graphics cards got left behind. The interesting part of that equation is that with the 8, 10, 12, 16, 24, and even 32 gigabyte cards they are going to quickly get outpaced by the next generation of models. A researcher that wants to use one of the models locally can probably be successful to get away with something above 16 gigabytes for now, but based on my usage of Claude Opus 4.8 the frontier models are going to quickly outpace what can be done with regular graphics cards. It’s an interesting situation where the next generation of models are not going to be able to run locally for most people.

Maybe what I’m trying to say here is that the development path forward using these types of agents is going to end up making code development abstracted going forward. I’m watching the development on GitHub and looking at what is being made, but the Hermes Agent needs very little help from me or input on the development choices being made. I’m more or less along for the ride and providing a few instructions, but I’m not the primary driver as the project continues in a very session-focused series of development efforts. Sure, we built a plan and worked on what would end up being worked on as a part of the process. The interesting part of the process was just how many tokens were used in the planning parts of the process. The overall spend on the Opus 4.8 model had these spikes for planning and then a more regular ongoing cost for the session development work. Opus 4.8 is amazing to work with. It is expensive to run, and I ended up watching the cost in tokens on a minute or hourly basis. During the start of this process, I decided that I was willing to pay for a couple of days of tokens to get a feel for how the development process with Hermes Agent and Opus 4.8 would work.

What exactly have I been working on during the last few days? I have worked on the Knowledge Reduce GitHub code that I had previously run about 10 different Manus sessions refining and working on, which has been published to GitHub over the last couple of years. I decided to take my most advanced work as a starting point and give the Hermes Agent my paper on the subject to refactor and produce a complete code package for that effort [1]. During the course of working with Manus, I had sessions that ran for hours on end and produced some interesting results, but none of that ever really turned the corner [2]. This current turn of the wheel with Opus 4.8 did manage to turn the corner, and the code that has been developed actually works now and can be verified by runs against my test data set, which is very interesting [3]. Assuming that you are willing to spend the token cost, the shift in what is now possible is very real, but it is very expensive to execute. The economics of token usage is now a very real factor in deciding what projects to work on and what coding efforts should be tackled based on cost.

Footnotes:

[1] https://github.com/nelslindahlx/KnowledgeReduce/tree/main/Paper_Preprint

[2] https://github.com/nelslindahlx/KnowledgeReduce

[3] https://github.com/nelslindahlx/Hermes-Output/tree/main/knowledgereduce

The initial download is pretty small, but it requires a bunch of installation and download. I don’t plan on giving it any keys or anything. During setup, the configure API keys and settings section was not really used. Right now I just want to kick the tires and see how it works running on my MacBook. I listened to the Intelligent Machines podcast this week, which made me question my donut choices, and it had Jeffrey Quesnelle as a guest [3]. You may know that name from the “YaRN: Efficient Context Window Extension of Large Language Models” research paper [4]. You may recall that back in 2023, this research helped push things from a context of about 4,000 words to 30,000 and opened the door to where we are now. That might not sound like much, but it was a huge unlock and a great open source contribution.

The download process was pretty easy, but I ended up subscribing to the base-level plan for $20 per month. We will see how fast the Hermes agent runs out of tokens, I guess. Things opened up, and I’m running Anthropic’s flagship Claude Opus 4.8 model to get things going. My first inquiry for Hermes was to find out if it can run a model locally instead of using cloud tokens. This got things started out by looking at my system and suggesting, based on my MacBook Air with 24 GB and an M3 chip setup, the Qwen2.5 14B model using Ollama to run locally. So far, things are moving along pretty quickly, and I’m starting to use the Hermes Agent without any trouble. The process has been pretty easy. At the moment, I’m just waiting for the ~9 gigabytes of the Qwen model to download. I have tried a few different models, and oh my goodness, did it make my MacBook Air run hot locally. This is the only time I have actually seen my fanless MacBook get hot.

Sadly, I spent my allotment credit during the first day. It seems like my best method to try out Hermes Agent on my MacBook is to use the cloud options and just accept that this experience is going to cost me a few extra bucks this month. I think running locally on my regular desktop computer would be fine. I have a good graphics card and 128 gigabytes of memory. I knew going into this that my MacBook Air was not the best device to test out a locally running model, but I was curious how it would end up going. My initial take is that the Hermes Agent is pretty powerful, but it really does want to have access to the machine, and so far, using the Nous portal models has made me less nervous about any key sharing. In terms of security, it seems like a better setup to reduce risk.

Footnotes:

[1] https://hermes-agent.nousresearch.com/desktop

[2] https://github.com/NousResearch/hermes-agent

[3] https://twit.tv/shows/intelligent-machines/episodes/874?autostart=false

[4] https://arxiv.org/abs/2309.00071 or https://arxiv.org/pdf/2309.00071 for the PDF

Yesterday, I ended up downloading Rufus to load up an ISO of Edubuntu on a USB stick. It turned out my old Dark Base computer had encountered some type of fatal error with a previous Ubuntu installation and needed to be refreshed. The new version of Edubuntu actually installed without any issues and even picked up the external wifi device without any issue. Finding all the right drivers is always the biggest concern with any Linux based installation. My old founder’s edition Nvidia card is still plugging along providing decent graphics. The older CPU and motherboard setup on that system won’t support TPM 2.0 and as a result Windows 11 is not an option, but Linux loads and works fine. This is the first time I had installed the Edubuntu flavor and so far I have been pleasantly surprised.

This weekend has been about recharging and getting refocused on some research efforts. Sometimes it is nice to just take a bit of time and avoid any real commitments for a couple days. Not having a plan was the right plan for this weekend. Taking a little bit of a break was really what I needed. I even had time to watch Star Wars Episode VII (2015) and the 7 episodes of The Book of Boba Fett (2021).

I’m still learning how to use Substack effectively after 5 years of being the platform. It feels like the social aspects of Substack are now more forward compared to the pure writing elements. Sometimes essays really do break out on Substack and go wide, but for most writers the process of building a larger community and grid of connections is what drives the conversation in a more deeply and ongoing way. I’m interested in zooming back out on all the AI we are seeing in our daily lives and the rise of the agent integrations is probably the right place to reconsider what we are getting out of the great AI bargains we are making. A lot of what people use and consume in terms of LLMs is just being passively used in search or maybe basic Copilot within the Microsoft ecosystem or Gemini within the Google ecosystem. A much smaller population of the AI user base is paying the larger subscription costs to code, build, or manage larger workflows. Right now the question and answer format of request and callback is still the largest use case within the entire AI space.

We are seeing the deeper coding based use cases pick up in speed and quality. It feels like Google is pivoting into competing with Anthropic’s Claude and other players have decided to put a lot of effort into advancing coding efforts. Those efforts are not going to end up being all you can eat free token buffets and we are going to see real limits or even caps on usage. Sam Altman has suggested it might go an entirely different direction and AI consumption will end up being metered like electricity or water [1]. Having a pay as you go framework for metered AI usage could be interesting, but at the same time some of these coding adventures could end up creating some very expensive bills within that cost framework. Certainly AWS found success in selling infrastructure and web services using very direct costing that avoided base monthly subscriptions. I’ll admit that I’m more willing to pay a small monthly fee to try out OpenAI’s ChatGPT or the Google AI Pro to utilize the Gemini ecosystem. It’s unlikely I’ll want to pay as I go with Sam Altman’s desired plan for my weekend coding projects. That seems like a path to ending up with a large unplanned bill and regret for letting a swarm of agents work on something that does not end up panning out.

Footnotes:

[1] https://www.tomsguide.com/ai/people-will-buy-intelligence-from-us-on-a-meter-chatgpts-ceo-sam-altman-has-critics-worried-with-his-ai-vision

This ongoing Functional Journal writing effort will continue to be 100% organic (devoid of token generated nonsense). For better or worse, these words will just be my efforts to collect my thoughts in written form on an ongoing basis. To that end, I’m going to publish more frequently under this domain. I’m not entirely sure when I’ll return to writing regular Lindahl Letter posts. That is probably going to work itself at some point. That writing project always seems to continue in some form or another. I’m over 200 missives deep within that ongoing narrative. Some of the older ones certainly contained some insightful nuggets. Writing weekly missives is something that has been a part of my routine for years. That will probably restart at some point. My goal here for the next little bit of time is to just write some daily posts and get back into the swing of effective word generation.

Part of that is sorting out a few things I have been considering deeply for the last few weeks. Namely how model distillation defenses are advancing and a way to recover the learning from all the models that are just languishing on Hugging Face and other repositories. The frontier models are super interesting and the token expense to train them is mind bogglingly expensive. All that sunk cost cannot currently be recovered. I had written a paper looking at how to advance a model framework that eats other models, but that work stalled out to some degree. I had coded up a framework to execute that type of work and had it running for a bit, but the results were not outstanding. It’s something I’m going to have to dig into more, but my attention has not been focused enough on it to move things forward.

Right now I’m interested in maintaining a cadence of weekly weblog posts where I share some observations and continue my practice of functional journaling. That is more or less something that has been ongoing for a couple decades at this point. I recently was considering what to do with that back catalog of content that for the most part is dormant online in archives and structurally inaccessible. At one point, I had pulled it all together into a corpus file that I could access by the ChatGPT 2.0 model they released. That was fun enough to chat with my previous writing and to train that model to write in my style to see how effective that would end up being. It was an interesting experience, but it eventually led to where we are at now which is a general and prolonged AI fatigue.

Let’s face it, the endless stream of banality creates fatigue. Reading great writing brings joy to people and has for a couple millennia. Being constantly exposed to middling and generic writing just makes taking in new knowledge and the journey of learning worse or more to the point perhaps it makes it tiresome. We are facing a new world of constant summary of a summary based content delivery where the original and novel fades away and sameness begins to be not an average of various inputs, but the constant drum beat of our generative model information age. An age that will probably remember more as a turning point in terms of how culture folded inward toward a repeating loop of normative fatigue.

Sure I’m watching season 13 episode 22 of the Curse of Oak Island right now while writing this missive, but I’m very confident that Lagina brothers are not spamming out AI generated content. They may have made 250+ episodes of searching an island in Nova Scotia for treasure, but it is most certainly not generated by an AI model. It’s really just an ongoing narrative of some people that came together with a shared purpose of searching for something with a relentless abandon. It’s probably that willingness to continue and strive forward that makes the show enjoyable.

Maybe the ongoing narrative that blogging provides is really about that same striving to edge forward by consideration of thoughts that appear within the forefront of the mind. Those top of mind topics can stack up and end up being just a lot to break down. Keeping things topical, but well considered takes a bit of diligence on the part of the writer. Making sure the content stays evergreen and meaningful for readers is an even harder bar to achieve. All that taken together is what separates some writers from others as they are able to build a backlog of excellence. A lot of my previous writing is just well previous and dated to the time when it was written. Some of it might veer off into the philosophical, but a lot of it is just a reflection of that moment that probably should just remain archived. It was a good and worthwhile activity to complete the consideration, but the output might be less valuable than the process used to create it.

Today we live in a world flooded with nonsense that is generated not curated by a thoughtful writer at a keyboard sharing an ongoing narrative with a reader. This is probably why podcasts are so successful. It’s a platform where a voice we know and over time grow to trust is presented and we know that the conversation is happening and was not generated. Sure, we are at the precipice of that changing wholesale, but the multi-hour conversation format will persist and the voices in that space that have gained traction by building fairly large audiences. Those audiences in some ways are larger independent media incarceration than the traditional broadcast formats that are now diminished. It’s my guess that the best of it will end up outpacing all of the other forms of communication.

At this point in our missive journey today, let’s circle back to the blogging community and how it has morphed, twisted, or just evolved into things like Substack and other newsletter delivery formats. People are engaging and finding communities of place, circumstance, and interest within these missives. Content written and shared probably still bubbles to the top and is given a better runway for success than generated content. Tokenized writing does not endure itself to a community of readers. My guiding north store for creating content here is to just write and let the writing be at the center of creating. I don’t really want to read generated content and to that end I’m not going to participate in publishing it online. Full stop.

I really have spent a lot of time thinking about how all the shared AI models published online could be combined, distilled, or just reused. So much effort went into creating all these models that have been shared on Hugging Face, but they are ultimately disposable in the current form. At some point, the big project of training and growing larger models may very well end up refocusing on gathering the very best of all the models into a grand model. That sounds fantastic, but I think it is actually something that will be possible. People have written about continuous distillation or data flywheels that maybe start down that path.

Google researchers have even published reports about increases in these types of attacks, which is really just a change in methodology [1]. Based on what they shared in terms of adversarial use of distillation, people are trying to figure out how to take a well-trained model and gain some advantage of continuous improvement outside of just training new models from large corpus efforts. Some of the largest corpus builds have actually come from book scanning projects. Google Books involved a huge scanning project, and Anthropic researchers followed a similar effort [2]. The team at Anthropic literally scanned millions of physical books to create a corpus of well-structured, gated, and known works. From those massive corpses models get trained just in the same way you can download an export from Wikipedia or The Pile to build a model.

Many people have built and trained models based on large corpus efforts. Based on the current state of the public internet, I’m not sure the quality remains the same as it was before. Something has changed, and quality just does not exist anymore. You could say that slop has destroyed the continuity of the written word. It is true in academics and in general online writing. I read a lot of academic papers, and you can tell that a lot of the current stuff is just not as poignant anymore. My argument would be that to stand on the shoulders of giants to advance the academy, papers have to make a substantial contribution to the field to stand out and become well-referenced and accepted within academia. Right now, the academic intake process is getting flooded with papers that don’t make a substantial contribution and are lesser academic works. That does not make the inquiry process any less important or the contributions of research any more valuable. It just means that finding the signal out of the noise is infinitely harder and going forward may be the permanent reality of the academy.

[1] https://cloud.google.com/blog/topics/threat-intelligence/distillation-experimentation-integration-ai-adversarial-use

[2] https://www.theverge.com/podcast/872998/anthropic-claude-books-netflix-theaters-vergecast

Sure, I started my day with a little bit of Farscape. The Olympic hockey games I want to watch don’t really kick off until Tuesday this week. It turned out that I had to subscribe to Peacock for the month to watch all the Olympic hockey games. It’s a small price to pay to watch the Olympics streaming, I guess, but I feel like they could have worked out some partner program to broadcast it without needing the monthly subscription. You can watch the main broadcast on NBC using the rabbit ears, but that is a broken system compared to streaming asynchronously the content you actually want to watch.

My winter Olympic watching is probably going to focus on hockey. The Colorado Avalanche hockey team has 8 players that will be competing. It should be good. I’ll admit that the 4 Nations Face-Off was actually really good television. I had no idea it would be such an enjoyable thing to watch in the world of sports. I don’t feel like the Olympic hockey coverage will get as much focus as that event based on the time difference and the volume of games. We will see how that plays out starting next week. This will be one of those things where I will try to avoid the scores so the games are still exciting during the replay.

We lock into a variety of adventures. Some of them are more defining than others. Today I’m sitting down and just waiting to lock into some focus time with football coverage in the background. My big plan for this block of time is to sit down and write. You know the old-fashioned long-form essay work that used to happen a lot more often. Recently, my online presence has been reduced. It makes sense. Our internet is not even ours anymore. What was the free and open internet where people shared things online has become silos of attention. It’s more of a one-way street than ever before. At first, websites showed up, and people went to them directly. Online was a space people arrived at like seeing billboards on the highway. Eventually, the spread of message boards and comment sections created some two-way conversations. At some point along the way, people engaged with each other, and streaming happened. Now we have a bunch of siloed islands of online content that are mostly streaming one direction with a few comment sections and chat interfaces.

Maybe I’m just nostalgic about what the internet used to be and won’t be ever again.

One of the things I have learned is that you need to spend time writing on your own the old-fashioned way. Sitting down and writing 1,000 words is a skill and something you have to practice on a regular basis. True stream-of-consciousness writing is not something that everybody can manifest on a regular basis. Sitting down and just writing has to be a part of your overall routine. Chatting endlessly with ChatGPT and Gemini is not the same type of writing effort. You have to bring your thoughts together and reason about the base form, function, structure, and assumptions of concepts to move beyond where you are at and begin to move beyond the edge of what is possible.

Here in a few minutes, the Colorado Avalanche are going to play a hockey game during the middle of the day on ABC with the widest possible audience. This season, I have been able to watch most of the games. It has been a great experience to watch the games this season. From the epic battles of the early season to strangely unsettling performances in the last few weeks. I’m sure things will get back on track.

Objective truth remains feeble these days. Slop abounds, and the basis of its creation is really just being explained. Models trained off the scraping of the internet at large certainly have a certain base of creation. People are certainly building alternate bases of thought, which can cause drift from a solid foundation to something else more self-serving. I spent some time writing about where the foundation of our knowledge base going forward will end up coming from as we go forward. One of the largest collections of information was the encyclopedias. You might be old enough to remember seeing a physical encyclopedia with the volumes separated by collections or letters of the alphabet. Now the largest encyclopedia is Wikipedia, and it probably has the most contributions from people of anything ever created. Seriously, no other collection of information has had so many contributors. We are now starting to see alternate versions show up where people are using large language models to generate mass rewrites.

That is a topic that is probably worthy of extended consideration. Maybe the question at hand is, “What is the foundation being used to feed the models going forward?” It is really a question of who controls the base-level information. The only thing that is probably bigger than Wikipedia is the massive knowledge graph that Google maintains. That thing is huge beyond what would have been possible before Google indexed everything. However, that big pile of everything that was online and indexed is now, for better or worse, breaking down as the internet that was seems to be coming apart at the seams. It’s possible that we reached the peak of what can be pulled together to train the base of these. I keep thinking back to one of the landmark collections of content that got scraped called The Pile which will always remain a sort of time capsule.

This was one of those weeks where my plans for productivity were rock solid and well articulated, but my actual writing output was subpar. That might be a bit of an understatement. My writing output for the week was lacking. I did, however, want to start playing with the new sensation setting my feed on fire this week, which is an open-source project called Clawdbot, which looks really interesting [1]. People seem to really like this thing and are even buying or setting up separate computers to run it all the time. I’m going to spend some more time digging into this one before I use any of the skill integrations [2]. The idea that people are just building and sharing all these skills for this open-source implementation is really interesting.

One of my biggest concerns of the whole agent race is just how powerful the integrations could become and how open they are to adversarial prompting and attacks. These types of ecosystems where people elect to connect more and more things to agents open the door to potential unintended consequences and bad actors. I’m sure it is one of the things that has slowed Google from allowing Gemini to just run wild across the broader ecosystem. One of the things I always consider the greatest potential risk vector is if people allow these things to make trades or take finance-related actions. Any mistakes within that type of finance workflow could be catastrophic. I’m going to try to set it up in a sandbox environment and not really connect it to anything impactful to kick the tires this week. I’ll let you know my thoughts next week about how that exercise ended up playing out.

Over the last few weeks, I have shifted to using Gemini more than ChatGPT, which has been interesting. I got really used to the ChatGPT application on my MacBook Air, and Gemini requires me to use a browser tab instead. I even went as far as to stop ChatGPT from loading at startup, which is a big shift in my usage. I was really curious about the personalized intelligence package they launched for Gemini [3]. We will see how that plays out going forward. Some of the personalization seems overly deferential and very repetitive in how it references concepts it wants to show you that it is aware of, which is off-putting. I write and work within a variety of topics, which seems to confuse the personalization they have so far running. Maybe they expect you to have a lot more deep work and focus on very specific things. I tend to research and jump from topic to topic each week. Like, for example, this week I’m very curious about that Clawdbot open-source project, but I may only be interested in it during the research phase.

Footnotes:

[1] https://github.com/clawdbot/clawdbot

[2] https://docs.clawd.bot/tools/skills

[3] https://blog.google/innovation-and-ai/products/gemini-app/personal-intelligence/

I almost sat down and spent some time working on a fiction project. It almost happened. That is the first time in some time that I wanted to sit down and just dive into creating a world with a written work of pure fiction. At one point, that was a lot more common in my writing routine. Maybe this weekend I’ll take inventory of the incomplete writing projects that maybe were a false start or received a lot of attention and then fizzled out. I know of a few of those projects that were 30,000 or more words into the process and then got shelved.

My Functional Journal weblog posts for the last few weeks have all used the same thumbnail and I’m okay with that going forward. It’s always an interesting process to pick a thumbnail for Substack posts. I’m not entirely sure how much picking the perfect thumbnail really matters. They really just show up on the homepage and when things are shared on social media. On platforms like YouTube the thumbnail is very important as it helps get people to click on the video. It is much easier to make a thumbnail these days. It’s really just a request process. My big plan for the weekend involves watching some football. The games this week should all be very good. It’s that time of year.

While the weather in Denver, Colorado is pretty cold it is otherwise nice outside. The Buffalo Bills came to town to play the Denver Broncos. Tickets were more expensive than I was willing to pay for a football game. I’m planning on watching the game OTA broadcast on CBS for no real cost outside of my time and electricity. Overall the people of Denver seem to be very excited about this home playoff game. Apparently, all that excitement was well placed and the Denver Broncos will get to play another game this year. The news after the game was disappointing, but it will make the next game very interesting.

Here is just a little bit I’m planning on going to Top Golf for a couple of hours which will be pretty fun. Hitting golf balls during the winter season is always a very interesting experience. However, the weather today will be pretty decent so the heat fans won’t have to do a tremendous amount of work to make the experience more comfortable. The high for the day is 51 degrees which is much better than freezing. It’s pretty warm for January in Denver, Colorado.

Originally it was just a weekly note published on Friday called the Lindahl Letter and it just brought forward things that caught my interest. That writing effort has yielded 220 weekly research notes on a variety of topics over the last 5 years. Almost exclusively the topics fall within the considerations of modernity’s shadow and the intersection of technology and modernity. Last year, I moved over my longtime weblog the Functional Journal to Substack and dropped all my previous writings within that effort offline. Some of that content can still be found on the Wayback Machine or other internet archives. To me, it was a fresh start and it meant that I had a forum for my technology based interests and a more reflective place to store these types of thoughts. Outside of those two main pillars of my ongoing writing efforts I did indeed stand up two additional writing efforts. One for Civic Honors and the other for nels.ai that are probably easily considered secondary writing efforts. At one point, I had thought maybe the nels.ai writing project would become my primary effort last year. I certainly focused in that direction and started producing daily content on November 3, 2025. Ultimately, that involved making some audio and eventually video content to further the endeavor. The Civic Honors effort really did remain secondary and it will probably only pick up occasional think pieces or maybe some future interview style posts.

Right now during this time of considering and ultimately choosing it is probably best to just focus on my two main writing pillars. That would involve producing a weekly weblog post for the Functional Journal that would be targeted for a Sunday release, but would be written throughout the week as either a series of thoughts or a larger essay. I’m comfortable being committed to ongoing weblog based writing and it as an effort should not distract from producing academic articles or research notes given that being a good reflective practitioner of the habit of writing requires that type of dedication. Beyond that weekly weblog post I have to figure out a way to transition my weekly Lindahl Letter publication into something more than just a couple of paragraphs and truly into a framework where research notes are developed that could become the basis for academic articles. At some level, I know intellectually that the entire academic article system is fundamentally broken. My hope is that at some point we will enter a new academic framework where the academy is renewed and making contributions to the greater academic good becomes easier and more streamlined.

At this point in my academic journey, it is distinctly possible that I can focus my research efforts and make my Lindahl Letter posts conform to my goal of producing academic articles in 2026. That is probably the best path forward at this point in time and something that I should commit to making a reality. Every year since finishing my dissertation it is something that I should have probably made a priority. We have the opportunity to strive forward and now is the time to make that a reality. Taking the first step to making this a reality will involve being committed to research and writing efforts in Overleaf using LaTeX to format papers for broader consideration and ultimately submission to journals. It really is just about making the decision to be committed to an ongoing academic process that yields papers.

Last week I took a week off from writing a weblog post. It was the end of the year and I just felt like it was time to rest and recover from the year. This week we are right back into the thick of things. We got some snow and high winds in Denver, Colorado during the winter holiday. We had such high winds that the gate and a couple of fence posts just blew over. Last week, I ended up having to buy a couple 4x4’s from the hardware store to replace fence posts. We got lucky that the weather was cooperating enough to allow pouring several bags of concrete in December. We had to get out the posthole digger and deal with some pretty chilly ground during the digging process. Ultimately, we built a new section of fence and the perimeter is now once again secure.

This is not the first time I have sat down and tried to write a 2025 reflection post. Some false starts did happen. Today I sat down to write out some end of year thoughts. Maybe I just don’t have any interest in recapping 2025. That is entirely possible. A bunch of people are probably just ready to move on 2026 and that is an okay outlook to have at this point. Maybe at some point in the future, I’ll be ready to write a more indepth reconsideration of 2025, but I just don’t see that happening any time soon.

I did spend a bunch of time (several hours) working on a paper called, “Post-Training Canonicalization via Continuous Multi-Teacher Distillation.” I even ended up with a draft of the paper over in Overleaf in full LaTeX format. I let the code for that one run for over an hour this morning and then sat down to watch the NFL Sunday morning pregame show on Fox using my OTA antenna. The broadcast signal looks pretty good and this is the primary way that I end up watching NFL football broadcasts. We pretty much watch whatever is on CBS or Fox and sometimes ABC. I’m going to circle back to the paper later this afternoon and read it again, but I need some time to pass to be able to pick it back up for deeper consideration. I need to figure out how to make a run pattern for the code that just uses 60 minutes of compute at a time so I can really dig into what is happening. Another alternative would be to buy some extra compute to test the code on a more powerful system.

All of my various writing backlogs have been moved over to Google Tasks and that has worked well enough. Instead of throwing things into Google Keep for later they are just getting stored in Google Tasks. Overall, that switch has not really impacted my workflow or the creative writing process. Part of that backlog may be just inherent in my backlog efforts is the assumption that I’m going to spend time learning something new. It’s maybe one of the more base things in my tree of expectations that it mostly just remains assumed. This week however the lack of a little bit of learning happening was enough to spark this missive. Last week I spent time thinking about the Substack video features and what it means to contribute video content. I even looked into how teleprompter video actually worked.

It’s probably time to do some rack and stacking within my backlog. That is something I should have some time to work on today. The remaining Kansas City Chiefs football games this season are mostly just background noise at this point. I plan on spending a few hours watching the Colorado Avalanche play the Minnesota Wild this evening assuming the streaming rights work out correctly. Sometimes even with the Altitude+ subscription a blackout happens from time to time. You would think that at some point you would just be able to buy the streaming rights to your favorite team for the season without needing half a dozen logins and services. The way fragmentation seems to be happening within the NFL broadcast rights that seems increasingly unlikely. Hockey and the NHL generally might have a better shot at just having everything on ESPN to lock into one service. My general view on the NFL is that I’m going to watch whatever games are on broadcast television using my antenna.

Strangely enough this morning the thing that really got me to lock in was my disappointment in the streaming practices of professional sports. That realization is somehow disappointing. It’s Sunday and I have time for pondering life, the universe, and well everything. While streaming sports rights are a very small slice of the everything part, it’s not a very important part of that consideration and all things considered the stronger part of my efforts should have been directed at something from my backlog.

For better or worse I have figured out how to get the Yeti X microphone setup to record and have tried a few different camera angles. It turns out that I can talk to the camera directly during a Nelscast type recording effort, but during the course of reading something I prefer a more indirect camera angle. The other thing I ended up doing was allowing the Yeti X microphone to be in the frame during the recording process. That gives the whole thing a bit more of a podcast vibe. I was able to publish video content with all 5 of the nels.ai posts this week and with the Lindahl Letter. That pace of writing and video recording is probably sustainable. I have really been enjoying the daily deep research into quantum, robotics, and AI/ML. Watching back my video content does remind me that I need to make sure to stay up temp and be consistent. At some point, I’m going to do a lengthy multi-hour Nelscat which will test my ability to deliver consistently as a content presenter.

Using the Substack video feature was helpful in terms of introducing simplicity to my workflow. Both of my 4K cameras work well enough, but the workflow from recording and editing video requires a lot more planning and effort. It is much easier to record a one take and go 1080P video from the browser. Maybe at some point I’ll end up switching back over to 4K quality video, but that switch is not something I’m planning any time soon. Maybe the most interesting part of this daily video posting experience last week was realizing that the YouTube shorts format picks up more views than anything else that I loaded. It seems like making videos targeted at the shorts format might be the way to go moving forward. Everything within this effort is all focused on my daily research notes which are driving my lifelong learning efforts to better understand technology and the world around me based on deep exploration.

During this missive we are talking about two very powerful elements of the creativity journey. First, being in the rhythm to write daily is an awesome super power and something I feel you can control through persistence and being open to the process of writing. Second, and more elusive by nurturing, hoping for, and recognizing the spark of creativity you can find something different and separate from that golden hour of writing. Maybe you can be set up for both of these types of moments through hard work and dedication. Having extreme preparation in a lot of areas helps have the foundation to deliver on the spark of creativity, but in the end is really just about recognizing the moment and seizing it so that no creative waste occurs. Maybe you were thinking about something and the outline of a book just presented itself or maybe the solution to an amazing argument that you could put into a treatise. For somebody who is more musically aligned you might just call it being in a flow space where the notes just show up and are most excellent. I’ll admit that even after years of playing guitar for me, those times where I just lock into a real flow space have become fewer and less prevalent. For me unlocking the superpower of creativity is more a writing based endeavor. I enjoy music, but ultimately that is for better or worse a more personal journey.

Earlier this week I started to evaluate what exactly I was using this writing super power to accomplish. Honestly, I know intellectually that I should be working building research notes into papers for publication. That is what needs doing and should be a part of my process. On a daily basis I’m taking the time to engage in research. During the weekend that tends to be for Lindahl Letter missives and during weekdays that effort is being folded into nels.ai which has been a focus of mine for the last few weeks. I’m actively sharing my research into the world of quantum computers on a daily basis. That work gets linked out to Substack Notes, Twitter, and Bluesky each and every weekday.

Facilitating that writing process for me is always about kicking the tires on my backlog. Earlier this week I dumped my writing backlog from a Google Doc into Google Tasks and I’m feeling like this was a brilliant decision. It took one big old cut and paste to just move the whole thing over to that new app moving a very long list to a lot of tasks. We will see how that works going forward as I try to complete tasks from the list. You have to invest the time to make sure the backlog is meaningful or your continued direction will diverge from the perfect possible future. This is an ongoing and eternal struggle for the path to be the right path. You can in fact within the Google Tasks app grab and organize tasks by stacking them which is how I define priority. Thanks at the top of the backlog are going to get looked at first, but that does not guarantee I will take action on them. It is possible they might get ignored in favor of some other item.

Thanksgiving is always an interesting week of travel and festivities. Today I’m digging into the process of writing with some early Thursday morning writing. Adventure awaits those who are willing to focus and put in the work. My weblog post from last week was a really good start to this trend of spending some focused time on the process of writing. It’s about opening a word processing document and spending an hour or more with the keyboard. I really enjoyed this piece from Mike Elgan about the internet needing a toggle button to turn off the AI slop and return the reader to the world of thoughtful well crafted prose. Not only did I elect to add that part about the prose being thoughtful and well crafted to the argument, but also be thankful dear reader that I did not elect to write a treatise about Luddites and the push against technology generally. That was a very real possibility.

I’m actually going to take the time right now to read the post from last week again and maybe give it a little bit of editing. Waiting a week between writing something and digging back into it gives me just enough time that I will catch those little missing words or slight typos. I’m not entirely sure how exactly those same slight typos become illusive in the moment of writing, but later become so obvious they are generally off putting. Oh they are so generally off putting. A few updates were made to that post from last week, but for the most part it was fine and that is the thing about revisiting some bit of past writing. Most of it really is just fine, but some of it needs a complete rework. I’m really into the idea of locking into writing for a set amount of time and just working at the keyboard until the words go from being forced into flowing. Sometimes that works out splendidly and other times it falters. Those times when it falters can be just as off putting as those now obvious typos, but maybe in a different, more frustrating way.

Within the Google Docs word processing environment you cannot just turn on the display word count feature to always be a part of the journey. During the course of typing on this MacBook Air I have to either hit Command + Shift + C or select “Tools” then “Word count” from the menu each and every time which is just tedious. Maybe some type of plugin for Chrome could be vibecoded that would always turn this word count feature on going forward. That is something I should probably add to my backlog. That task has actually been added to the backlog. We will see what ends up happening with that one here in the next couple of weeks. I sat down here to write 1,000 words to kickstart the day into super awesome mode. Instead, I have edited a couple of documents and finished off my breakfast of Huel ready to drink nutritionally complete protein shake.

Today I spent about an hour cleaning up all the old posts from my LinkedIn profile. It was not something that needed to be done or was really worthwhile, but darn it if I did not feel accomplished after completing that post clean out task. I’m thinking about adding that to my task list just to be able to cross it off, but that seems ridiculous.

Beyond writing with confidence you have to be willing to put in the work. I think that is the key part of the equation that these new agent interfaces like ChatGPT are fundamentally changing. People find it easier cognitively to allow the machine to produce and to work with the output vs. sitting down and writing something more comprehensive. I have felt that pressure as well recently. I’m just not willing to sit down and write for 3 hours at a clip anymore. I just sit down and lock in for 1,500 words, maybe 2,000 and then I’m done. A lot of the time my efforts just end up being a quick 500-750 word draft and then I move along to something else. My efforts to do that are interesting and they fit into the 4 major writing buckets that I have been focused on maintaining. However, I struggle to figure out if the time was spent in a way that moves things forward toward that perfect possible future. My interest in AI/ML, AGI, robotics, and quantum are all very real and active, but that research in the aggregate is being amassed toward what end goal remains a strong and real question.

It feels like we are going to get back into the habit of producing some of these weblog missives that have sections and are not one endless string of prose. That is probably for the better. It is also the result of my decision to try to guide this type of writing efforts into one big essay each week that is targeted to be shared on Sunday. It’s a weekend writing effort to bring it all together and put it into a package that is sharable, approachable, and ultimately shares the main point even if that point happens to wander around for a bit before becoming apparent.

Engaging in some paper writing

After attending my 20 year college reunion at the University of Kansas last week I was thinking about engaging in some paper writing. More to the point, I was thinking about how to reset my paper writing efforts to somewhere around the transition from college to the professional world. My advisors at the time of graduation did not push me to collaborate on anything in particular or publish papers from my dissertation. I took my dissertation results out to a few conferences and started down the journey toward academic efforts. Things after that just took another direction. It might now be possible to ask Gemini 3 (which is surprisingly good at the deep research type things I like to do) or ChatGPT to engage with me to facilitate writing a couple of papers. That is one way to go about it or I could perhaps engage in conversations with other academics that are writing at a regular clip and try to triangulate the right path forward toward being productive.

Sustained writing efforts

Over the last few weeks or maybe months at this point, I have engaged in writing efforts typically from start to finish on a topic very quickly. Normally, I’ll circle back to what was written for a quick bit of editing and then whatever it is gets published directly or scheduled. I’m not really engaging in sustained writing efforts that last multiple hours. That is something that I certainly could spend time working on and it might be something that is very worthwhile. Maybe toward the end of the year I can take my revised backlog and really spend some time in deep work. I’m going to try to complete some 3 hour writing sessions. That will be very different from my 30-45 minutes of focus that I’m currently achieving. Keep in mind that my current strategy works and I’m able to produce decent research notes, but they fall short of longer manuscripts or paper level efforts. That is something that I feel like is lacking in my day to day productivity and it is something that has bubbled up to the forefront of my thoughts. It’s something obviously that I am concerned about.