orchestrator -> parallel subagents with investigation, authoring, verification, benchmarking subagents and integration / final verification handled by parent has improved my productivity too.

I feel like from here its agent swarms against a whole spec but haven't got there yet.

Still getting plenty of bugs in the more complex scenarios, but mostly (in some projects) i never have to look at the code and treat it like a black box

This grew out of a larger post where I try to explain diffusion models in less-dense terms, especially the parts that felt non-intuitive to me when first reading the papers. That post emphasizes that diffusion models incrementally “step” toward a solution rather than generating it all at once, and this interactive zooms in on that sampler-step idea specifically.

Would love feedback on:

- whether the interaction makes the sampler update easier to reason about

- what other diffusion concepts would be useful to turn into small interactives

Link: https://www.bryanthornbury.com/interactives/sampler-step-exp...

Comments URL: https://news.ycombinator.com/item?id=47200769

Points: 2

# Comments: 0

For many (including myself) who haven’t stretched those particular math muscles since diff eq class a decade or so ago, the paper is just an opaque wall of literal Greek.

In this post I describe my personal understanding of diffusion models in less-dense terms, focusing on intuitive understanding and personal mental models I use to understand diffusion.

Comments URL: https://news.ycombinator.com/item?id=47200240

Points: 2

# Comments: 1

Lots and lots of tests!

In another view, standard libraries do a pretty good job.

https://arxiv.org/pdf/1804.03720

For day-to-day work it is far superior than endless virtualenvs clogging up my harddrive and hunting for brew package dependencies.

I open sourced a template in case other may find it useful.

                      {
   name:         "Large Input Slice",
   input:        []any{"A", "B", "C", "D", "E", "F"},
   chunkSize:    3,
   expectedChks: [][]any{{"A", "B", "C"}, {"D", "E", "F"}},
  },
  {
   name:         "Remaindered Large Input Slice",
   input:        []any{"W", "X", "Y", "Z", "1", "2"},
   chunkSize:    4,
   expectedChks: [][]any{{"W", "X", "Y", "Z"}, {"1", "2"}},
  },

I have almost completely replaced the time I usually spend searching stackoverflow for the simple, routine stuff (like how to easily filter a list in golang like I can in python), and then adapting whatever answer I find with a simple query to Mixtral.

Usually, Mixtral gives me a direct answer to my question that requires no actual adaptation.

However, I found, frequently the code is not quite correct and seems to be a blend of programming languages.

To get around this, I started asking Mixtral for a test for whatever code it generates, and this was a game changer. To state it simply, the tests are pretty good, and require very little tweaking to get running. With the tests I have confidence in the code it generates.

Once, I even asked it to write a test for one of my own functions and it said it could do that but first I needed to know about a bug in the code, AND IT WAS RIGHT.

All of these little ventures to stackoverflow add up and I am convinced that using Mixtral has saved me one or two days just for this project that took about a week. For a project where I was more familiar with the programming language, I'm not sure the time saving would be so dramatic, but I'm impressed nonetheless.

Now, this post wouldn't be complete without mentioning that it hasn't all been roses. A couple times it needlessly complicated the code, and once it was even wrong (trying to parse scientific notation), but it still saved me a ton of time.

Here's one of the tests it generated:

    func TestChunkSlice(t *testing.T) {
        testCases := []struct {
            name         string
            input        []any
            chunkSize    int
            expectedChks [][]any
        }{
            {
                name:         "Empty Input Slice",
                input:        []any{},
                chunkSize:    3,
                expectedChks: nil,
            },
            {
                name:         "Input Slice With Zero Chunk Size",
                input:        []any{1, 2, 3},
                chunkSize:    0,
                expectedChks: nil,
            },
            {
                name:         "Single Element Input Slice",
                input:        []any{1},
                chunkSize:    1,
                expectedChks: [][]any{{1}},
            },
            {
                name:         "Small Input Slice",
                input:        []any{1, 2, 3, 4, 5},
                chunkSize:    2,
                expectedChks: [][]any{{1, 2}, {3, 4}, {5}},
            },
        }
    
        for _, tc := range testCases {
            t.Run(tc.name, func(t *testing.T) {
                actualChks := ChunkSlice(tc.input, tc.chunkSize)
    
                if len(actualChks) != len(tc.expectedChks) {
                    t.Fatalf("Expected %d chunks but got %d chunks.", len(tc.expectedChks), len(actualChks))
                }
    
                for idx, exp := range tc.expectedChks {
                    if len(exp) != len(actualChks[idx]) {
                        t.Errorf("Expected chunk #%d to have %d items but found %d items.", idx, len(exp), len(actualChks[idx]))
                    } else {
                        for jdx, val := range exp {
                            if !reflect.DeepEqual(val, actualChks[idx][jdx]) {
                                t.Errorf("Mismatch between expected (%v) and actual (%v)", val, actualChks[idx][jdx])
                            }
                        }
                    }
                }
            })
        }
 }

Comments URL: https://news.ycombinator.com/item?id=39751674

Points: 3

# Comments: 2

Comments URL: https://news.ycombinator.com/item?id=32613816

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=25853230

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=21287679

Points: 2

# Comments: 0