Early Version

Yarvin is blogging tonight, so we are co-blogging. His article is titled A New Sovereign Accounting.

Yarvin wants to maintain the American Empire in Washington, and I want to eliminate it. We have several obstacles to doing so. The nature of the currency and the nature of the military are the most prominent ones.

His sovereign accounting starts with three rules:

• The government owns the country. This is an unusual type of view of government, but it's not an unknown view. - The individuals in the state are mainly part of the collective of the state. If they oppose the state to a large degree, they should be liquidated, either killed or relieved of their assets and power.
• Fiat currency is government equity. This is a new rule that seems to be Yarvin's own: fiat currency is a proxy for specie currency, such as gold and silver. We are almost out of living memory for gold and silver being currency. But in the absence of government, people look to the laws of nature and of art. And of natural scarcity to create value.
• Securities are formal. He seems to misunderstand what a security is here. He defines a contract guaranteed by law to be a type of security. 💡 I would define a security as a financial instrument structured in a regular way, publicly registered, and based in the interest in some real property. ⚔️ this definition is intended to exclude derivatives; they are not securities

Yarvin thus introduces two contradictions and errors. Dollars are not a share of the U.S. government, or at least they should not be. The current policy of 2% per year inflation cannot go on forever.

Yarvin does not yet discuss the stocks and flows problem. That is, does a wealthy individual need to spend money to gain disproportionate influence, or does simply having the money bestow such influence? If you imagine dollars as "shares", it would be the latter.

🔥 If you imagine dollars as "shares" in the US government, it will be owned by China soon.

Click title to read full message...

There were 4 California OAH hearings this month.

The first case ⚙️ Case No. 2024100129 Parent v. San Diego USD. The topic was a very long, approximately 90-minute bus ride for a student with autism who had to be restrained on the ride. The verdict was that a maximum of one hour would be reasonable for the 12-mile drive in San Diego County.

The second case ⚙️ Case No. 2024100512 was Parent v. Fresno, USD. Student was represented by parent. The case was 17 pages and not very interesting. None of the parents' claims were found to have merit.

The third case ⚙️ Case No. 20214101050 seemed an exercise in bureaucracy. 🔥 We are unable to comment publicly on counsel in the case. The remaining claim was a request by the school for due process hearing to have the court authorize the IEP. The parents had elected not to consent to the IEP. After a three-day hearing, the court, in a 56-page ruling, determined the IEP was reasonable and allowed it to be implemented.

The fourth case ⚙️ Case No. 2024110856 concerned a student who was paralyzed and did not have control of their bowels. The question was what nursing care is required and whether this needed to be discussed and litigated in an IEP meeting. The verdict was that the provisions for the student's medical care were insufficient.

from https://jasonpargin.substack.com/p/to-save-the-world-save-yourself :

I see a certain sense that people want it both ways regarding Special Education.

As I see it, there should be a binary distinction.

Either: you are in the general education classroom.

Or: you are in Special Education.

----

The problem is that we have parents who, on the one hand, request Special Education services ... but, on the other hand, demand placement in the General Education classroom.

These are both mandated by the IDEA. They are contradictory mandates.

----

For many students, there is not a need for full-time Special Education. The student can spend 80% of their time in General Education, and 20% in Special Education. ⚙️ both accelerated / honors courses and remedial courses would fit this pattern

For other students, there is a need for full-time Special Education.

----

When there is a conflict regarding the time spent in Special Education, the default outcome should be that no time is spent in the General Education classroom, if there is a colorable argument that the student's presence harms other student's educational outcomes. 🔥 this raises a perilous question: if it is found that the presence of a disruptive / violent student in a classroom was beneficial to other students' educational outcomes, would it be incumbent on the State to allow or engineer such situations?

Huggingface had approximately one open-source model that supported Lithuanian TTS.

Fortunately, one model is all that is needed.

----

It is called "OuteTTS" ⚙️ Huggingface https://huggingface.co/OuteAI/Llama-OuteTTS-1.0-1B ; website https://outeai.com .

And, after a bit of handholding 💡 and a sample audio file from OpenAI.fm to generate a "speaker profile", it works. All the letters are pronounced.

----

It is a local model. Which means that it will take a few hours to generate 🔥 checks notes 723 words/phrases to read. ⚔️ well, actually, 2169 if I keep doing three different voices for it

there is a solution to this.

have the phrase be "Pradžia : pirštinės : pabaiga" (begin, WORD=gloves, end).

And then, match the wave-forms to trim the audio file.

----

that is: if you look at a .WAV file, you can see clear patterns based on the letter.

the FFT will say the overtones.

💡 I never did build AUDIOLAB.

⚙️ Faulkton SD is along US 212, west of Watertown.

Today's task: finding a speech-to-text model that supports Lithuanian.

The choices seem to be OpenAI.

----

ElevenLabs supports: English (USA, UK, Australia, Canada), Japanese, Chinese, German, Hindi, French (France, Canada), Korean, Portuguese (Brazil, Portugal), Italian, Spanish (Spain, Mexico), Indonesian, Dutch, Turkish, Filipino, Polish, Swedish, Bulgarian, Romanian, Arabic (Saudi Arabia, UAE), Czech, Greek, Finnish, Croatian, Malay, Slovak, Danish, Tamil, Ukrainian & Russian.

Google Gemini supports: Arabic (Egyptian), German (Germany), English (US), Spanish (US), French (France), Hindi (India), Indonesian (Indonesia), Italian (Italy), Japanese (Japan), Korean (Korea), Portuguese (Brazil), Russian (Russia), Dutch (Netherlands), Polish (Poland), Thai (Thailand), Turkish (Turkey), Vietnamese (Vietnam), Romanian (Romania), Ukrainian (Ukraine), Bengali (Bangladesh), English (India), Marathi (India), Tamil (India), Telugu (India)

----

PlayAI ⚙️ https://play.ht seems to support Lithuanian.

Also, Coqui AI ⚙️ an open-source project; https://github.com/coqui-ai/TTS claims to support Lithuanian.

----

The main problem with OpenAI is that, sometimes, the words are truncated.

This has been most common with pirštinės in testing; the leading P sound is missing.

the tool of the day is Replit.

----

In what may be a good business decision, but is slightly annoying to the consumer, one must pay the $25 (monthly) up-front before using Replit.

----

Replit uses Claude; so one cannot expect it to be particularly better or worse than other tools that also use Claude.

One advantage is that the "billing" system is very transparent. It was 5c per "change" for the initial changes I made. ✨ if you aren't paying for it, in the long run it doesn't exist

One disadvantage is that there is a very low limit for the maximum "context" one can pass into queries. It was, generally, three files. This is rarely enough.

----

It also creates Git changes in a way that almost requires a development repository, and "flattening" before pushing to main 🔥 "main" used to be called "master", but the language police decided they should change it.

----

The UX changes it made were a clear case of "one step forward, one step back". Which ... can be helpful. But, the constraints of Replit make it less helpful.

The second "LLM coding app" tried was Windsurf.

And, while it is not a knockout decision, ZenCoder has clearly won on points.

----

The biggest advantage of Windsurf over ZenCoder is the usage dashboard. It is much clearer how many queries I have run, and what they have cost.

----

The coding skill of Windsurf was substantially worse.

Was this because it defaulted to their own (free) SWE-1-Lite model, rather than a more effective one? Maybe.

----

Both have a version of the "accept changes" diff-review tool. Windsurf lets you accept part of the change. But, in practice, this is more of an inconvenience than a convenience. When the change is something like "update the imports", there are 8 separate things to accept.

----

Vibecoding is still slow enough that I can operate two sessions simultaneously. So, perhaps, having Windsurf and Zencoder both running can double my productivity.

----

Zencoder's "advanced workflows" do seem to have solved some of the basic problems. Like figure out which file the user is referring to. The solution is not "read everything", but "use a subcommand to grep the file titles".

----

Ultimately, for an experienced coder, the key feature of these tools is to fail in a way that is easy to detect and correct.

⚙️ https://muse.jhu.edu/pub/1/article/922346

They Don’t Read Very Well : A Study of the Reading Comprehension Skills of English Majors at Two Midwestern Universities

----

And, beyond the fold, Bleak House

Click title to read full message...