Information from Documents @docs2info

Read documents. Not too many. Mostly PDFs. Our parents read and wrote documents. Having the world's knowledge an internet connection away doesn't change this. docs2info.com Santa Rosa Joined September 2021

Tweets

2K
Followers

397
Following

374
Likes

6K

Information from Documents @docs2info

2 weeks ago

I haven't enjoyed an AI talk as much as this one for a long time. A smart person doing practical work. Just chipping away at a problem intelligently and diligently. youtube.com/watch?v=pxPCKR…

0 0 3 98 1

Emmanuel Ameisen @mlpowered

3 weeks ago

youtube.com/watch?v=fGKNUv…

1 2 22 2K 22

(((ل()(ل() 'yoav))))👾 @yoavgo

3 weeks ago

in this work, we show that reasoning texts are really not transparent, in the sense that we (humans) are really bad in inferring how the steps (really) depend on each other for the model. yoo can test yourself here: do-you-understand-ai.com

Mosh Levy @mosh_levy

3 weeks ago

8 27 136 26K 98

Download Image

2 14 98 11K 60

Information from Documents @docs2info

3 weeks ago

The greatest joy modern developers experience is putting their, or their employers' or customers' data into a proprietary online service. LLM foundation model, data store etc. linkedin.com/pulse/githubs-… They do this from laptops that are more powerful than 1980s super-computers.

0 0 0 40 0

Mosh Levy @mosh_levy

3 weeks ago

Producing reasoning texts boosts the capabilities of AI models, but do we humans correctly understand these texts? Our latest research suggests that we do not. This highlights a new angle on the "Are they transparent?" debate: they might be, but we misinterpret them. 🧵

8 27 136 26K 98

Download Image

Information from Documents @docs2info

3 weeks ago

I still can't believe people took scree plots seriously, let alone reverentially. youtu.be/geUH6K8o9nw?si…

0 0 1 54 0

Lance Fortnow @fortnow

3 weeks ago

I vibe coded a Turing Machine accepting 0^n1^n for the Theory of Computing class I'm teaching this fall. inquisitive-lolly-1bb788.netlify.app Took me about 10 minutes on bolt.new. This is just too easy.

1 2 10 2K 2

Jack Morris @jxmnop

4 weeks ago

the OCR conjecture: some examples include artifacts such as OCRV ROOT, which indicate the training data may have been reading between the lines: OpenAI is scanning books (for some reason the model loves mentioning how many deaf people live in Malaysia)

8 11 657 67K 62

Download Image

Hamel Husain @HamelHusain

a month ago

> sign up for new AI coding product > it’s free, get an email asking setup a meeting to onboard My brother, this is the most expensive software imaginable.

9 5 106 11K 6

Rohan Paul @rohanpaul_ai

a month ago

Swedish Prime Minister is using AI models "quite often" at his job. He says he uses it get a "second opinion" and asks questions such as "what have others done?" At the moment he is not uploading any documents. IMO, when these models are capable of giving seemingly better…

95 82 486 252K 127

Download Image

Information from Documents @docs2info

a month ago

This is how one of the big LLM vendors' chatbot told me to avoid big LLM vendor lockin. docs.google.com/document/d/1Jw… "Build an Abstraction Layer This is your single most powerful defense." This is the advice our parents gave about pre-AI vendor lockin years ago.

0 0 2 63 1

Steve Krouse @stevekrouse

a month ago

Vibe code is legacy code @karpathy coined vibe coding as a kind of AI-assisted coding where you "forget that the code even exists" We already have a phrase for code that nobody understands: legacy code Legacy code is universally despised, and for good reason. But why? You have…

91 148 948 234K 482

Download Image

Information from Documents @docs2info

a month ago

The go file editing agent example is impressive. github.com/johndocs/agent… shows a convoluted way to write a javascript fizzbuzz.

Information from Documents @docs2info

a month ago

The go file editing agent example is impressive. github.com/johndocs/agent… shows a convoluted way to write a javascript fizzbuzz.

0 0 0 195 0

0 0 0 101 0

Information from Documents @docs2info

a month ago

A simple example of the variation between runs of a simple Claude 3.7 Sonnet code summarization from the Go agent sample. github.com/johndocs/agent… For some reason, the variation has reduced dramatically. With so many factors to logged, I missed variation.

Information from Documents @docs2info

a month ago

0 0 1 185 0

0 0 0 195 0

Information from Documents @docs2info

a month ago

There's so much variation between runs on this simple agent that I struggle to see how people are going to use libraries like this in commercial products.

Information from Documents @docs2info

2 months ago

There's so much variation between runs on this simple agent that I struggle to see how people are going to use libraries like this in commercial products.

0 0 2 159 0

0 0 1 185 0

Information from Documents @docs2info

2 months ago

Simple Agent code in Go github.com/johndocs/agent copied from ampcode.com/how-to-build-a…

0 0 2 159 0

Information from Documents @docs2info

2 months ago

Some coders, like DHH, learn through their fingers. I never have. I think about problems, write out long hand summaries, draw pictures, build prototypes, read etc then write out draft solutions.

ℏεsam @Hesamation

2 months ago

Some coders, like DHH, learn through their fingers. I never have. I think about problems, write out long hand summaries, draw pictures, build prototypes, read etc then write out draft solutions.