amboo7

A->A, B->C is also known as the Fibonacci law of hiring: 1st class people hire 1st class people (but may also hire 2nd class people), 2nd class people hire 3rd class people, etc.

1. none 2. nvidia nim, opencode zen, google free models. local QFN iq3_xss on m1 max 64gb, paperniuk splash 3. pi agent, pi pocket 4. no 5. no

You should have tried Deezer. I had switched to Spotify because I couldn't tolerate bugs anymore.

Agreed. For the others, I asked Claude to write a structural edit tool: s-expressions only, insert/replace/... When passed 'tree' it prints a file like

<line>: <s-expr .-separated path>

then eg when passed 'replace 2.1 "(+ x y)"' it returns a new file (as a list of s-exprs).

"180° wrong" as "0° right"? Just that bit, not that I disagree.

Me too

I have my own. One Telegram bot for reading, summarizing etc. (using a local Qwen 3.6), another for improving it (OpenCode Telegram, free DeepSeek V4). I might put it on github soon.

I have always liked coding (started in the 80s). But I got bored having to deal with trivialities (spending life on fixing commas) and increasing accidental complexity (having to memorize tons of APIs). Now LLMs deal with that crap and I can apply stuff I read in research papers and such, with much lower threshold.

Trying to recall...

# 27B GGUF # Benchmark results: # - TG speed: ~5 tok/s (vs baseline ~4.5 tok/s, +10% improvement) # - PP speed: ~60 tok/s (stable across context sizes) # - With parallel=4: total throughput ~28 tok/s

llama-server \ -m /Users/*/models/hf/models--unsloth--Qwen3.6-27B-GGUF/snapshots/82d411acf4a06cfb8d9b073a5211bf410bfc29bf/Qwen3.6-27B-Q6_K.gguf \ --alias "qwen3.6-27b" \ -ngl -1 \ --n-cpu-moe 0 \ -fa off \ -ctk q4_0 \ -ctv f16 \ -c 131071 \ -b 512 \ -ub 256 \ --spec-type ngram-cache \ --jinja \ --cache-ram -1 \ --parallel 4 \ --kv-unified \ --no-context-shift \ --mlock \ --slot-save-path ~/qwen_slots \ --reasoning-budget 512 \...

# 27B MTP GGUF # Benchmark results (from bench_conversation.sh, ~500tok prompts + multi-turn): # Config | PP (tok/s) | TG (tok/s) | Draft accept # ---------------------------------------|------------|------------|------------- # Baseline (ngram-cache, non-MTP) | 48.0 | 5.2 | N/A # MTP --no-mmap, dn=4, fa on, ctk q4_0 | 45.7 | 6.9 | 60% # MTP --no-mmap, dn=3, fa on, ctk q4_0 | 46.3 | 7.5 | 73% ← winner

MODEL="/Users/bale/models/hf/models--unsloth--Qwen3.6-27B-MTP-GGUF/snapshots/ac393bc3d23fd5a929a85e2f33c7c4fd5be02d43/Qwen3.6-27B-Q6_K.gguf"

llama-server \ -m "$MODEL" \ --alias "qwen3.6-27b-mtp" \ -ngl -1 \ --n-cpu-moe 0 \ --no-mmap \ -fa on \ -ctk q4_0 \ -ctv f16 \ -c 131071 \ -b 512 \ -ub 256 \ --spec-type draft-mtp \ --spec-draft-n-max 3 \ --jinja \ --cache-ram -1 \ --parallel 1 \ --kv-unified \ --no-context-shift \ --slot-save-path ~/qwen_slots \ --reasoning-budget 512 \...

# 35B GGUF # Benchmark results: # - TG speed: ~27 tok/s (vs baseline ~21 tok/s, +30% improvement) # - PP speed: ~370-350 tok/s (slight decrease with larger context) # - Parallel=4 gives best throughput at ~152 tok/s total

llama-server \ -m /Users/*/models/hf/models--unsloth--Qwen3.6-35B-A3B-GGUF/snapshots/9280dd353ab587157920d5bd391ada414d84e552/Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf \ --alias "qwen3.6-35b" \ -ngl -1 \ --n-cpu-moe 0 \ -fa on \ -ctk f16 \ -ctv f16 \ -c 262144 \ -b 2048 \ -ub 512 \ --spec-type ngram-cache \ --jinja \ --cache-ram -1 \ --parallel 4 \ --kv-unified \ --no-context-shift \ --mlock \ --threads 4 \ --threads-batch 8 \ --slot-save-path ~/qwen_slots \ --reasoning-budget 512 \...

# 35B MTP GGUF # Benchmark results (128K context, verified 2026-05-18): # TG: 30.7 tok/s, Draft accept: 65%, no OOM at 128K

MODEL="/Users/**/models/hf/models--unsloth--Qwen3.6-35B-A3B-MTP-GGUF/snapshots/e28512781649329c5b37cbf55029355a48d158d4/Qwen3.6-35B-A3B-UD-Q6_K_XL.gguf"

export GGML_METAL_BF16_DISABLE=1

llama-server \ -m "$MODEL" \ --alias "qwen3.6-35b-mtp" \ -ngl -1 \ --n-cpu-moe 0 \ --no-mmap \ -fa on \ -ctk f16 \ -ctv f16 \ -c 262144 \ -b 512 \ -ub 256 \ --spec-type draft-mtp \ --spec-draft-n-max 3 \ --jinja \ --cache-ram -1 \ --parallel 1 \ --no-context-shift \ --threads 4 \ --threads-batch 8 \ --slot-save-path ~/qwen_slots \ --reasoning-budget 512 \...

Very cool. Anything publicly available?

Same happened to me. I only had to ask the service where to tap exactly.

For me the combination of agentic search and deep wiki creation for legacy code bases is a killer app/game changer. Not for the surviving authors of the legacy code bases.

Whay about of the tons of caches that just pile up until you notice that you must delete them manually?