Message Chain

⚙️ Fort McKay is located about 500 miles north of Calgary.

🤖 Fort McKay is a small community in northern Alberta, located about 54 kilometers north of Fort McMurray along the Athabasca River. It has historical roots as a Hudson’s Bay Company trading post and is now situated in the heart of Alberta's oil sands region. The local economy benefits significantly from nearby oil sands development, with businesses and partnerships tied to the industry, though environmental concerns remain a key consideration. Surrounded by boreal forest, the area is rich in natural beauty, and traditional activities such as hunting and fishing are still part of the community’s lifestyle. Fort McKay offers essential services, including a school, health center, and recreational facilities, while more specialized services are accessed in Fort McMurray.

goals for the week, "professional":

• new personal website (with resume/portfolio; and also with photographs and "list of hobbies")
• Atacama / LLM integrations
• Atacama print-view tuning
• Atacama "small feature" list: database re-structuring will be at the top of the list

goals for the week, "personal":

• social events on Monday and Thursday
• get haircut 💡( not today, not Sunday; the barber shops in central Iowa are closed or overcrowded on Sundays)
• additional withheld from public view

There are some directional choices to be made about Atacama. Will I be importing content from other platforms? Is it going to support multiple users and/or multiple installations?

Some of the answers will be based around how the LLM features work. And look.

https://thehill.com/homenews/senate/5066601-senate-democrats-consider-rfk-jr/

Some Democrats said to be open to backing RFK Jr.

💡 I mostly agree with the take. RFK Jr is controversial in different ways than most of the other controversial nominees of the next administration.

Over the past 2 months, the stock for Rigetti Computing (NASDAQ:RGTI) has gone from 1.12 to 19.02 . 🔥( and yet, quantum computing is still fake)

I must assume that Google's "error-correction" paper is the driving factor. It was at 1.35 on November 20 when Google announced "AlphaQubit". ⚙️( https://blog.google/technology/google-deepmind/alphaqubit-quantum-error-correction/ ) 💡( AlphaQubit is nothing. It is an LLM that does a computational task very slightly better than a much simpler model), and 4.4 on December 9 when Google announced Willow ⚙️( https://research.google/blog/making-quantum-error-correction-work/ ) 🔥( Willow could be Big If True. )

Of course, quantum computing is still fake. 💡( even with Shor's algorithm, one cannot accomplish anything useful with 49 bits in 58 microseconds) But, that isn't the question for the markets. The logic is "LLMs are big, quantum computing feel similar, QED".

The question is: when will the Samaritans get off the train? I have no answers.

football is done.

⚔️ well, actually, there is still "week 18" of the NFL, plus two more rounds of college football playoffs, and ... does the East-West game count?

it is college basketball season. much like football has done, there is too much supply for me to care about any individual game right now.

unlike in football (where Iowa State is always the "scrappy underdog" or the "lovable loser" or "team with a great year wins the Pop-Tarts bowl"), ISU sometimes is legitimately good in basketball. 💡( both men's and women's; although this year only the men's team is possibly "legitimately good") this might be one of those years.

four of the top 6 teams in the polls are in the SEC. the other two are Duke and ISU.

https://www.espn.com/mens-college-basketball/story/_/page/CBBblockbuster0103/sec-men-college-basketball-dominance-auburn-tennessee-florida

ESPN describes how good the SEC has been.

although, I have one gripe. If you look at the team records (as-of 5 January), the worst home record in the conference is 8-1. The most road games any team has played is 3. 💡( this is not unique to the SEC; most major-conference basketball programs play 5 games against bad teams at home in the non-conference. it lets them sell tickets, it lets them get easy wins, and it hopefully doesn't erode the integrity of the sport too much.) ⚔️( well, actually, some teams have as many as six "neutral-site games")

goals for today:

• write the long Atacama to-do list
• laundry
• Peloton ride 🌎( the attempts to use a level to level the bike were not successful. but I think it is good enough to ride.)
• shopping?
• no terminal/IDE usage until after sunset

There are 26 things on the Atacama to-do list I wrote today. That doesn't count the "multi-player features" that have been asked for.

💡 it is possible that the full list will show up in a different message.

separately, there is "log into an old computer with my AWS credentials, and see if there is anything of value (other than email archives) still there".

one game of chess tonight, against Maia (1100 rating). 🔥( it hung a bishop on move 3)

I missed one move, but it missed it as well. beyond that, it was 45 moves of slowly suffocating the position.

I spent a total of 6.5 minutes on my moves, and also got to think for about 5 minutes on the opponent's moves. ⚙️( being able to efficiently use your opponent's clock time is a key chess skill) 💡( the careful observer will note this is more than the roughly 3 minutes one gets in 2+1 chess)

in the NFL: Week 18 is meaningless. several teams were probably throwing ⚙️( deliberately trying to lose), either for draft position or to avoid having to play Cincinnati in the playoffs.

it would be improper for the players on the field to try to lose. but, keeping your eight best players out of the game is accepted.

In The News: Justin Trudeau expected to resign as Prime Minister of Canada. This will not immediately trigger an election; the next planned election is in October.

🔥 the machine won't know yet why he is resigning. ChatGPT, after a search, says it is "internal party pressure" due to "policy disagreements", which says nothing.

💡 as far as I can tell, "everybody is tired of his schtick" and "Trump" might be the only two reasons.

⚔️ well, actually, the Freeland resignation ... is not a root cause. it is normal politics once a situation is over-determined.

⚙️ the infamous "every government from the COVID period is doing badly in elections" trend is also relevant.

⚔️ well, actually, i can't do "one line per bugfix" here. there is too much interference.

So, here are the things from the 26-item list I expect to finish before the next dispatch.

• Fix sigil boxes in dark mode.
• Fix "rotating sigils" - an idea half-created by the machine.
• Fix single-message page - there is some old/bad CSS (no "block" for colortext, scrollable second column)
• Create "read more from stream" page
• Fix font color in "Style picker" widget
• When an unsupported method is used (such as POST /), don't throw an exception
• Create robots.txt
• Create favicon.ico
• Create apple-touch-icon.png
• Improve server logging

If all that is done, I will start looking at the other 16 items.

also it snowed a lot in various places.

there is no other news since last night.

approximately all of the tasks in 3/12 are done. at the end, I hit Claude's "please come back in a few hours" threshold, with a paid account.

🔥 after seeing "rotated sigils" implemented, i turned it back off.

for the afternoon, some Atacama tasks of a different complexity-class.

💡 presumably at least one of these will turn out to be a bad idea and/or take more time than it is worth

• stop calling them "Emails" in the DB schema. possibly migrate to a Dispatch / Frame structure (so sections of emails can be filtered separately)
• print-mode tuning. better placement for two-column mode, smaller text, less padding / line space, color check. 💡( generating a PDF would be nice, but is dependent on other things that aren't done)
• add an option for "login-required" when writing posts; these shouldn't show up in the stream
• add plumbing for LLM calls. calls to GPT-4o-mini are cheap enough (and competent enough) to be sufficient.
• update the landing page to be less passive-aggressively useless
• the new lexer, again

the details for these will probably be more interesting than "I asked Claude to do it, it kinda-mostly did it, I cleaned it up". and in a separate reply.

also, non-Atacama tasks.

• as mentioned earlier, "clean AWS"
• get a new personal website landing page
• 20 minute Peloton bike ride 🌎( the wobble factor was fixed by tightening all the screws)
• probably two chess matches, 10+10, against humans online

https://www.sourcenotes.blog/p/5-ideas-to-carry-into-2025

💡 this whole "you can't punch down" meme has reached the end of its useful life. there is too much perverse anti-status signaling going on.

The targets in this conflict are the elites—those perceived as sitting at the top of the social hierarchy. Gioia lists wealthy CEOs, DC politicians, celebrity TV newscasters, law enforcement authorities, experts of all stripes, Ivy League academics, movie stars, and other figures of authority or expertise. “All of the cultural energy right now is on the bottom. And that energy has been intensifying,” he observed.

https://www.infinitescroll.us/p/weekly-scroll-meta-swallows-the-slop

🔥 the LLM takes aren't terribly edifying, but the soup content is good.

I wish there was a name for the common failure mode I see in modern culture - it goes something like this: 1. There are some highly autistic people that have real issues with certain foods or certain behaviors can’t just adapt or move on. 2. There are some people who self-diagnose or claim to be ‘autistic’ when really they’re just entitled and selfish and they enjoy being weird jerks. 3. Society has become more aware of the first group and more accepting of the first group over time, but this has also led to explosive growth in the second group, who are now empowered to be pricks about anything they want and claim ‘autism’.

💡 it is literally the same problem as the previous frame. weaponization of "I'm low-status so you can't criticize me" by people who are objectively high-status.

🔥 poor people don't whine if they can't have a $47 bowl of soup

the machine succeeded at none of the more-complicated tasks I had planned for yesterday afternoon.

today there is no Atacama work. tomorrow there will be some, but it will be triaged.

the main goal is "new personal website". ⚔️( well, actually, some content from Atacama might show up there.)

the chess and the biking were good, mostly; i needed an ibuprofen this morning. 💡( which means both "I need more biking" and "I need no biking today")

In The News: Facebook tacks into the winds of the Trump administration.

some of these things (moving content-moderation teams to Texas) are probably things they already wanted to do, and now have a good excuse for.

others are kow-towing.

but, the headline is that they are "getting rid of fact-checkers". 💡( for reasons that survive scrutiny)

There are several massive problems with the c.2021 fact-checking regime. For one, the fact-checkers definitely have veered into "liberal opinions are true and other opinions are not, regardless of the facts". 💬( They said they were researching misinformation, not researching preventing misinformation. -- Dylan Alvarez) 💡( Ben Collins, one of the most prominent "misinformation" pundits, now runs Fake News site The Onion.)

But, more importantly, the problem is: "by doing enough fact-checking, you imply that everything that survives fact-checking is true". Which was never the case.

And, more broadly, the concept of "people vote on what content is popular" is inherently flawed. Facebook works as "what your friends are talking about", but it is as bad as every other site when filled with "viral" content.

🔥 aren't viruses a bad thing?

it will not be a 12-part email next week.

the rhythm of 2 emails per day does not work for me.

💡 "save message when the browser restarts" is not currently a feature.

the new https://www.pow3.com is up.

there are various letters to be written:

• a cover letter on what I can do, what I want to do, why I haven't been doing it for the past few years, etc.
• on LLMs in general, and my LLM coding experience. 💡( there are a lot of "unnecessary" commits because of the need to pass changes back-and-forth between my own branches)
• the LLM benchmarking data from November/December
• investments I have made (and will advertise)?

also the mailserver is closer to working.

i played two games of chess today. one win, one loss.

chess is a game about hoping the other player makes a mistake. against perfect play, you cannot hope to win. 🔥( and tu cannot hope to even draw.)

features for Atacama for tomorrow:

• The "frame-dispatch" schema migration. 💡( that is: instead of just generically calling them "emails", there will be a dispatch, separated into various frames) ⚔️( well, actually, I don't know where the "chain" logic will be. having multiple replies is useful. but, a "main branch" is also useful.) ⚙️( the hr tags separate the dispatch into frames.)

It is likely that some other useful feature will arise naturally out of the migration. ✨( tomorrow's details will be described tomorrow and not today )⚔️( "getting tests running in PRESUBMIT" is orthogonal to the database schema.)💡( some version of "the LLM writes a summary if all the text is in colortext blocks" is the most likely feature)

surely one of the USB drives around here has a Linux install drive on it?

💡 "get the c.2016 laptop running Linux, because Windows on an HDD is too slow" probably won't be done today.

I have come to the conclusion that, while the machine is good at some coding projects, it is uniformly bad 💡( or, at least, not good) at sysadmin tasks. It's not useless for them, the results for specific questions are still better than a Google search. But if you ask it to do a mildly complex task, it will almost certainly get something wrong. 💡( there is no reason to think this is any "universal truth"; it is more a reflection of the available training data and the limitations of an LLM-only system)

today's primary goal is the "frame-dispatch" database schema migration.

there are subgoals:

• set up the PRESUBMIT check for Python imports
• ensure the local environment can create the alembic schema migration scripts
• add some additional database fields for desired features 🔥( will they be "hashtags" or "channels"? or "topics"?)

I'm not stating any LLM-specific goals with the schema migration. But it is possible either "LLM-generated topics" or "LLM-generated copy-edits" or something will suddenly become very easy.

non-Atacama goals:

• time-permitting, today is the day I should get a haircut. and, go shopping while i am out.
• nothing else remarkable is planned.

well, most of the print-view changes are done.

the Substack Notes feed seems to be unusually terrible this week. Many of the posts are several months old. There is too much "bro-sphere" dating advice posts. There are three different posts about "porn star Lily Phillips fucks 100 men in one day". Some casual stereotyping based on nationality. And, of course, more apologetics for Luigi Mangione.

🔥 does that say something about the targeted user? / not really, I have already blocked a lot of this content on Substack.

plans for today:

• additional UX polish based on user feedback
• "portfolio" / "recent experience" section of personal homepage
• some additional writing tasks
• social event, scheduled at 5PM