symbolicAGI

A NYC bank has given its coding agents email addresses.

Learned from an NYC exchange 10 years ago that Java can be written so as to not use garbage collection. Fast and no pause for GC.

1. Resource and reuse objects that otherwise are garbage collected. Use `new` sparingly.

2. Avoid Java idioms that create garbage, e.g. for (String s : strings) {...}, substitute with (int i = 0, strings_len = strings.length(), i < strings_len) { String s = strings[i]; ...}

I launch Claude Code in VS Studio (similar to Cursor): > claude

Then I use the /login command that opens a browser window to log into Claude Max.

You can confirm Claude Max billing going forward in VS Studio/Claude Code: /cost

"With your Claude Max subscription, no need to monitor cost — your subscription includes Claude Code usage"

Given chat and API needs for low-latency, llama.cpp is probably still the best choice for self hosted models with or without GPU support. And Ollama is the leader for wrapping llama.cpp.

Because Tokasaurus was mentioned as better than Ollama for conducting darwinian godel machine operations (self-improvement), I looked for the linked repo on GitHub and it was 404. So glad it is back https://github.com/ScalingIntelligence/tokasaurus.

AI First company should have that role be a set of AI agents. Founder must provide the overview of tasks mentioned in the post.

Evidence that the path to ASI is not extending the capabilities of LLMs, but instead distilling out and compiling self-improving algorithms externally in a symbolic application.

On the other, asking the LLM to generate the source and test code for a poker game would be more productive. Could actually be a better player than some people with an AI algorithm for decison making.

Radiation protection indicates the life form is space-born.

Guess it is on Mars also - whenever we get there.

Fascinated by spiders and insects growing up in Upstate NY - the largest jumping spider there gets 20mm long. Their eyesight and reflexes are fast enough to stalk a landed house fly and catch it on its takeoff.

Still feel comfortable today in a deep squat from those days long ago.

GPT 4o-mini is 1000x less expensive than o1-pro and works well for coding tasks when examples are given in the context.

Plus if your application can run queries in batches with high latency, then OpenAI batch API offers half price fees.

There is sufficient circumstantial evidence for bacterial life on Mars.

The (perhaps religiously) conservative science establishment refuses to conclusively accept it.

When a robotic mission is sent to Mars with the express purpose of finding current life, then it will become obvious that hardened bacteria spores not only survive in our solar system's space, but thrive in the most difficult environments. And what about that stuff growing outside the international space station?

The question about why certain bacteria on earth are very resistant to the sun's harmful radiation is waiting to be solved.

The threat to some of the world's religions is the hypothesis that the simplest life did not originate on this planet first. So what?

It matters and the author's guidance for mitigating future immense risk is welcome.

Marxist theory will bend to encompass the total automation of Amazon fulfilment centers. Expect future marxist arguments for UBI.

Pull requests are too-downstream - a legacy I believe.

An AGI system that self-improves its code will regenerate every component impacted by the enhancement starting from live system design narratives, useful existing components, relevant design patterns, and intermediate development artifacts that are discarded or become stale in human-driven legacy coding.

I see "agents" as mostly bodies of assembled prompts for LLMs of various strengths used at the appropriate time in the pipeline of code development. A code review agent's prompt would not have the task of generating code and thus not need all that particular context, but would look for historically observed 'gotchas' and flag those for automatic repair, and the repair could go all the way back through the artifact chain to the text requirements and specifications.

One nit-pick with the article:

It links to a Forbes article arguing that the best chess is played by humans and computers working together.

This concept is evidently false. Alpha Zero played super human chess without any human assistance beyond coding its machine learning algorithm and inputting the rules of chess.

Furthermore, in a timed chess match the delay and possible misjudgments from a human would detract from the performance of the human-computer team versus the computer chess program alone.

This well-thought article restricts the notion of code review to pull request assessment - diff file - on an existing code base, perhaps a mostly human-written legacy code base.

As a Java code generation researcher, I would appreciate any constructive comments regarding the automated code review of automated changes made to an entirely generated code base, whose live specifications and requirements have been modified in order to generate the code being reviewed.

Starlink provides high bandwidth plus low latency for international connections. The low latency occurs because once the origin TCP packets reach the low orbit Starlink network, they can be almost single-hopped to the destination continent before downlinked to the user's Starlink antenna. Land based routers take 10 ms or so per hop and there are many on an international connection.

The White House property Wifi dead spots would be best improved separately by a commodity secured mesh network.

For example, Google Fiber Austin to Israel (land based router hops): 1 _gateway (192.168.1.1) 1.162 ms 1.155 ms 1.191 ms 2 [redacted] 3.077 ms 3.054 ms 3.035 ms 3 * * * 4 * * * 5 * * * 6 ae1.pr02.dfw101.googlefiber.net (192.119.17.142) 8.666 ms 7.731 ms 7.698 ms 7 * * 141.101.74.42 (141.101.74.42) 9.144 ms 8 * * * 9 * * 104.18.28.18 (104.18.28.18) 9.160 ms

Cisco Systems declined 89% from its dot-com peak to its bottom during the crash.

Expect the same for Nvidia.

The huge datacenter orders for redundant infrastructure will start postponing and cancelling this year. That is a falsifiable prediction.

Because the AGI system's alignment with ideal human values entails behaving in a friendly helpful manner.

One might reasonably ask a frontier how to generate the source code for an agent based system that exhibits examples of initiative, emotion, creativity, curiosity, opinions, beliefs, self-reflection, or even logical reasoning.

In the USA, Republicans build the military and Democrats wield it.

Makes sense because the mass of the known universe intuitively is enough to prevent even light from escaping its total gravity.

Should our universe accordingly be spherical?

Can we detect outside objects entering the universe's event horizon?

What if any statements can now be made about what goes on inside the black holes that we can observe?

Git is a block chain in which successive data entries (commits) include the hash of the predecessor in the data such that the entire data set can be verified against data corruption by recomputing the hash codes of each successive entry and comparing the final hash code with the separately recorded latest hash code for the data set.

Other sequential data sets use this efficient consistency check, for example the Kafka log files do this.

In contrast, cryptocurrencies for example Bitcoin, use block chains whose hash codes are crypographically (slowly) calculated so that it is practically impossible to corrupt the data to arrive at a given hash code.

Git and Kafka use very efficient hash codes that do not have this anti-hacker feature. They only detect ordinary corruption from example missing, duplicated or garbled data as opposed to malicious data falsification.

"Blockchain" in common speech has the meaning of cryptographic block chain and that is why Git is not ordinarily considered to be a "Blockchain" despite it having data blocks whose verification is performed by chained (efficient non-cryptographic) hash functions.

Used cannabis for 55 years. Switched to temperature controlled vaping for 10 years to avoid inhaling benzene and other toxic vapors from the combustion of plant material.

Now using only edible forms of cannabis to totally avoid inhalation of toxic vapors.

Wondering if any studies exist relevant to consuming edible THC and CBD?

Back in the 1960s United States, women were often perceived as more sensitive to public profanity, compared to men of the same age.

MIT AI Lab back in the 1960s published technical reports containing program code.

The military slang 'FUBAR' f'ed up beyond all recognition, was in the student and professor engineering vocabulary. The tradition became to use 'fu' and 'bar' as nominal function names, in same manner as X and Y were nominal variables.

Often in the MIT technical reports, one would see 'x = fu(y)' or 'y > bar(z)' and so forth. If you knew, you knew.

A few years later, perhaps with the welcome progress of more female faculty and students, textbooks changed the spelling, but not the pronunciation of the vulgar acronym 'fu' to 'foo'. Again, if you knew, you knew.

And now you all know.

With my LLM written Java code I prompt for null checks using assert statements during one of the improvement iterations. Any reference variable that could be unexpectedly null is followed by a null check immediately after its assignment.

Given that Java source code will increasingly be written by AI, it makes sense to incorporate coding patterns that are both defensive and clear for an AI to read.

Additionally, LLM code generation understands and writes Java using null values to represent cases of object absence.

"Canagliflozin extended the lifespan of mice with premature aging even when treatment was started in middle age."

Canagliflozin is an existing prescription drug, with possible side-effects: https://www.mayoclinic.org/drugs-supplements/canagliflozin-o...

A previous study showed a gender effect: Canagliflozin extends life span in genetically heterogeneous male (14%) but not female mice (https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7710304/)

NASA and EU should consider the option of sending the required laboratory to Mars in a one-way trip, even with a rover to bring the cached samples back to the landed lab.

What aspects of a life-determining laboratory must be present on earth that cannot be hardened for use on Mars?

No need to expensively quarantine samples on Earth, when the analysis is performed on Mars and the digital results sent back to Earth. The presence of fungal and bacterial life on Mars would wonderfully disrupt so-what thinking about exploring our solar system.

Better for NASA to consider the fastest and cheapest way to get the results from life-detecting analysis of the cached samples on Mars.