ljosifov

I usually doubt the 'small dataset tuned' variants. B/c ages ago (in the NN prehistory) I've done some NN training, and appreciate how hard it is to improve in general, and how easy it is to ruin a model in general while targeting a small dataset (LoRA-s are ok, that's different). That model/quant was the most recent one I was trying. But could not really use any of them, as the combo model + llama-server ground to a halt even at small context depth sizes on the amd gpu.

Yesterday I finally found a good combo! So writing this for the benefit for anyone that may have the same h/w. Got around to /GOAL search for something better for the h/w (amd 7900xtx), and pi agent found a new best that actually seems it will be useful for real. As the 40 tok/s speed starts dropping only at 260K context depth?? Served by hipfire from this repo https://github.com/Kaden-Schutt/hipfire, that worked the best got on llama-benchy:

  | Context | Wall Time | PP (t/s) | TG (t/s) |
  |:-------:|:---------:|:--------:|:--------:|
  |   1,024 |   17.3s.  |   660    |    40.8  |
  |   8,192 |   29.3s   |   606    |    40.8  |
  |  32,768 |   70.5s   |   599    |    40.7  |
  |  65,536 |  126.5s   |   594    |    40.6  |
  | 131,072 |  235.5s   |   591    |    39.0  |
  | 196,608 |  349.0s   |   591    |    40.2  |
  | 260,000 |  451.3s   |   594    |    37.2  |

  hipfire pull qwen3.6:27b           # 14 GB MQ4 quant, qwen3.6-27b.mq4
  hipfire pull qwen3.6:27b-draft     # 0.92 GB DFlash draft, qwen36-27b-dflash-mq4.hfq

Hobbled - but not to death, the few times I use it (usually on a plane). I tried 2bit of a 20% REAP reduced experts. :-O That's the biggest that fits on my own h/w (3yrs old M2 Max 96gb). It's coherent, it does work, doesn't fall apart on casual use. IDK if better than dense 27b. Think 27b was slower on the same h/w. DS4F has got 1M context window. Nowadays with weeks long run hermes sessions, I get to 300k-400k context depths easily. The speed decline profile of DS4F with context depth increase is superior to any other model I try. (I try them all - love this stuff) Only previous model coming close on that is nemotron-cascade-2 (only 30b-a3b) - that also has 1M context window.

True - they are workhorses. Not super bright, but good enough for lots of everyday tasks. I've found sweet spot to be turning thinking off, as it adds small or no value, while increasing the token count and waiting time. Last 27B I used was https://huggingface.co/Jackrong/Qwopus3.6-27B-Coder-GGUF - specifically post-train adapted a bit to run with thinking off. I saw today the 35B-A3B MoE from the same HF acc is out, downloading that rn to try.

Running 27B dense model on M5 128GB is ok, but one can do better.

On M5 128GB one can make use of the ram and use sparse MoE. For example, DeepSeek-V4-Flash will fit, served by DwarfStar (https://github.com/antirez/ds4). One will probably improve 2x the token/sec speed, given DS4F 13B activated params in the MoE are ~1/2 of the ~27B of the dense Qwen.

27B Of the Qwen fit even on a cheaper 24GB card, e.g. amd 7900xtx (<$1K?) or slightly dearer nvidia 3090 (with cuda). With ~900 GB/s bandwidth they will likely be ~50% faster than the M5 with 600 GB/s.

Haha :-) - FoxPro and Clipper next.

Not replaced but supplemented. For off-line coding current setup is pi + ds4-server + DeepSeek-V4-Flash REAP25 (on M2 Max 96gb). For simpler programming related (e.g. text2sql) as well as synthetic data generation, current best for me is llama.cpp + Gemma-4-26B-A4B (on gpu 7900xtx 24gb; sometimes nemotron-cascade-2-30b-a3b for 1M context). That and (dabbling now) auto-research uses lots of tokens. Used to get paused running out of token quotas all the time. The 1st local model I found somewhat useful to me was glm-4.7-flash, and it's gotten way better since. Recently between OpenCode Go choice of models at many price points, and DeepSeek-V4 dropping the IQ/$$$ by multiples, have become less reliant on local llms for this auxiliary work. Claude I use but with Zai GLM-5.2 subscription. And maintain GPT subscription for quality models.

For high Ram (unified), and relatively middling to lowish Tflops and bandwidth GB/s, usually MoEs are most hopeful. The current top-1 in the (iq, tok/s, @ context depth) ranks for me (M2 Max, 96gb) is DeepSeek-V4-Flash REAP25 <65gb gguf + ds4-server + pi agent. Not better than cloud API ofc, but useful enough to endure if I need to. E.g on a non-Internet 4h flight the battery (local llm draws 60w) held long enough. REAP supporting ds4 branch here

https://github.com/ljubomirj/ds4/tree/reap-compact-support

DS4F dropping to unusable <10 tok/s only at 784K context (!!) makes a big difference.

Yes, it's performant, and esp performant at non-trivial context depths. DeepSeek-V4 DS4 (and Flash - DS4F) drop tok/s speed much less than the rest. On my M2 Max it took context depths of 768K to drop tok/s to ~10 tok/s.

https://x.com/ljupc0/status/2062457314414587996

Other local models I've checked drop to unusable speeds way sooner. Only other model with similarity favourable curve I've tried is nemotron-cascade-2-30b-a3b. But it's a small model, way dumber than DS4F.

Coding agents use cases have large context depths. The rate of decline is as important as the headline number.

Thanks for the tear down. IDK anything about quantum (my knowledge there starts and ends with https://www.scottaaronson.com/democritus/lec9.html), but amused enough to follow in the background. See whether it ends crazy-bad or crazy-good. :-)

And just today now I saw this https://arxiv.org/abs/2606.07352. What's your take on it?

Am old enough to have witnessed more than one wave of "impossible things" happen for real in my lifetime, so lets see. As long as the scientific method (evidence, publication, replication, testing, etc) is mostly followed am curious enough to check blog posts or interviews from time to time.

+1 for boring. Boring code is Solid Code, in the sense of "Writing Solid Code" - the old book by Steve Maguire.

Thanks for the DS4, will give it a try. Was hoping maybe I can re-quantise shave few GB... MiniMax-M2.7 Unsloth's UD-IQ2_XXS is down to 65GB - it run albeit too slow to be usable to an agent at context depth. I'm curious DS4F with it being economical with the KV caches - if that translates into keeping up with context. Was hoping 80GB 2-bit quants maybe come down to 70GB... that would be more comfortable to run.

On 96gb I can give up to about 88GB to the GPU with sysctl iogpu.wired_limit_mb=88000, without suffering any ill-effects. When pushed higher I tend to notice e.g. graphic driver errors, youtube web page not working, other semi-random glitches. So the ~80 GB of DS4-flash quants I could just about fit. Leaving some extra for the KV caches. Will try, I'm curious how's the DS4 degradation with context depth growth, how fast does tok/s drop. E.g. 2-bit lowest quant MiniMax-M2.6 runs, but starts low tok/s and degrades fast with context depth.

The biggest models I can comfortably run are about 1/2 the DS4F size - like gpt-oss-120b. Lately was toying with Ling-2.6-flash. Got the agents to adapt existing metal kernels in llama.cpp, and it did run (model https://huggingface.co/ljupco/Ling-2.6-flash-GGUF, branch https://github.com/ljubomirj/llama.cpp/tree/LJ-Ling-2.6-flas...). It's 104B-A7B4, and for the M2 Max 7.4B active is about the most it can take while still producing 40 tok/s. And the hybrid arch allows for graceful degradation, still close to 30 tok/s at 64K context depth.

Too bad L2.6F while the best have, is not that much better in agentic benchmarks compared to my current incumbent local llm (nemotron-cascade-2). Got inspired by DS4 to start a l26f branch (WIP https://github.com/ljubomirj/l26f). :-) Try squeeze the most from L2.6F. There should be low hanging fruit in good integration of the agent and the inferencing engine. On input - considering the huge difference cached v.s. non-cached tokens. On output - considering that the NN gives us the complete logits set for all 200K+ tokens vocabulary.

Love this, even if can't use it atm (not got the h/w - only 96gb on M2 Max). I get it the general comp/public will find it unusable or worse. Reminds me of how home computers were - mere toys - before they became personal computers (PC). On my h/w the only passable combo for me atm is pi agent + llama.cpp + nemotron cascade-2 model: to 1M context, hybrid arch doesn't crash & burn 1/N^2 with context depths of 10K-50K-100K used by code agents. Was on a plane without Internet the other day. Brought a smile to my face that I could run pi agent (with llama.cpp serving), and it was just about usable at 40-30 tok/s. Afaik the usual API speeds are double that, 60-80 tok/s. Sensors showing using 60W when running inference. So battery probably would not last more than >3h. Model only 30B in size leaves plenty of space for KV-caches, and other programs - even at generous 8-bit quant. Only 3B active params at one time (with MoE A3B) is about the most that ageing M2 Max can carry it seems.

What we see and experience - it's all natural, it's the natural order. :-) When people claim something is un-natural, usually it's natural in that occurs in nature, only - they themselves find it objectionable. It's something they personally dislike, and would rather it not happen.

In the same boat with 7900xtx. 24GB vram, on paper decent performance, in reality most things don't run. Only llama.cpp is consistent that it can run most models, even if maybe not at top performance (afaik - lacking MTP, problems cache invalidation with hybrid models). At least with llama.cpp I know what runs. With various python-based inferencers, between their uv/venv, my venv, system envs/pythons/libs yadayada - I need an agent to get to the bottom of what's actually running. :-) Yeah IK skill issue/user errors - but don't have seconds in the day left to spend them on that.

Even if not perfect, if you publish on GH or HF, some other agent can maybe start there and not from zero. I did this for Ling-2.6-flash (107B-A7B4 MoE) that's the biggest llm I can ran for practical use on the other h/w I got for local llms (M2 Max). Even if MTP is not working well, still improvement on the current llama.cpp that does not run Ling-2.6-flash at all. This - https://huggingface.co/inclusionAI/Ling-2.6-flash/discussion.... The 4-bit quants are at https://huggingface.co/ljupco/Ling-2.6-flash-GGUF, the branch is at https://github.com/ljubomirj/llama.cpp/tree/LJ-Ling-2.6-flas....

~/llama.cpp$ build-.../bin/llama-batched-bench -m models/....gguf -npp 512,1024,2048,4096,8192,16384,32768 -ntg 128 -npl 1 -c 36000

  On amd 7900xtx

  Qwen3.6-27B-Q4_K_M
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.743 |   689.35 |    4.605 |    27.80 |    5.348 |   119.68 |
  |  1024 |    128 |    1 |   1152 |    1.188 |   862.17 |    4.573 |    27.99 |    5.761 |   199.96 |
  |  2048 |    128 |    1 |   2176 |    2.566 |   798.09 |    4.602 |    27.81 |    7.168 |   303.57 |
  |  4096 |    128 |    1 |   4224 |    5.936 |   690.00 |    4.639 |    27.59 |   10.575 |   399.43 |
  |  8192 |    128 |    1 |   8320 |   15.034 |   544.90 |    4.729 |    27.06 |   19.763 |   420.98 |
  | 16384 |    128 |    1 |  16512 |   42.807 |   382.74 |    4.886 |    26.20 |   47.694 |   346.21 |
  | 32768 |    128 |    1 |  32896 |  137.377 |   238.53 |    5.188 |    24.67 |  142.566 |   230.74 |

  Qwen3.6-27B-IQ4_NL
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.535 |   957.45 |    3.715 |    34.45 |    4.250 |   150.59 |
  |  1024 |    128 |    1 |   1152 |    1.124 |   911.16 |    3.677 |    34.81 |    4.801 |   239.97 |
  |  2048 |    128 |    1 |   2176 |    2.447 |   836.89 |    3.698 |    34.62 |    6.145 |   354.13 |
  |  4096 |    128 |    1 |   4224 |    5.711 |   717.17 |    3.729 |    34.32 |    9.441 |   447.43 |
  |  8192 |    128 |    1 |   8320 |   14.615 |   560.52 |    3.821 |    33.50 |   18.436 |   451.30 |
  | 16384 |    128 |    1 |  16512 |   41.966 |   390.41 |    3.967 |    32.26 |   45.933 |   359.48 |
  | 32768 |    128 |    1 |  32896 |  135.789 |   241.32 |    4.253 |    30.09 |  140.042 |   234.90 |

  On mbp M2 Max

  Qwen3.6-27B-UD-Q8_K_XL
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    2.583 |   198.18 |   22.049 |     5.81 |   24.633 |    25.98 |
  |  1024 |    128 |    1 |   1152 |    8.321 |   123.06 |   22.364 |     5.72 |   30.685 |    37.54 |
  |  2048 |    128 |    1 |   2176 |   17.873 |   114.59 |   23.290 |     5.50 |   41.164 |    52.86 |
  |  4096 |    128 |    1 |   4224 |   41.967 |    97.60 |   23.624 |     5.42 |   65.591 |    64.40 |
  |  8192 |    128 |    1 |   8320 |   68.722 |   119.20 |   21.077 |     6.07 |   89.799 |    92.65 |
  | 16384 |    128 |    1 |  16512 |  142.184 |   115.23 |   22.026 |     5.81 |  164.210 |   100.55 |
  | 32768 |    128 |    1 |  32896 |  339.778 |    96.44 |   24.465 |     5.23 |  364.243 |    90.31 |

  Compared to similar prior models

  On amd 7900xtx

  Qwen3.6-35B-A3B-UD-Q4_K_S
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.203 |  2517.60 |    1.482 |    86.35 |    1.686 |   379.67 |
  |  1024 |    128 |    1 |   1152 |    0.427 |  2399.22 |    1.471 |    87.04 |    1.897 |   607.15 |
  |  2048 |    128 |    1 |   2176 |    0.946 |  2165.23 |    1.478 |    86.59 |    2.424 |   897.67 |
  |  4096 |    128 |    1 |   4224 |    2.253 |  1818.33 |    1.502 |    85.22 |    3.755 |  1125.01 |
  |  8192 |    128 |    1 |   8320 |    5.849 |  1400.51 |    1.525 |    83.91 |    7.375 |  1128.17 |
  | 16384 |    128 |    1 |  16512 |   17.115 |   957.27 |    1.589 |    80.55 |   18.705 |   882.78 |
  | 32768 |    128 |    1 |  32896 |   56.008 |   585.06 |    1.704 |    75.10 |   57.712 |   570.00 |

  Qwen3.6-35B-A3B-UD-IQ4_XS
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.204 |  2508.94 |    1.313 |    97.46 |    1.517 |   421.78 |
  |  1024 |    128 |    1 |   1152 |    0.423 |  2418.64 |    1.296 |    98.80 |    1.719 |   670.18 |
  |  2048 |    128 |    1 |   2176 |    0.946 |  2164.61 |    1.323 |    96.78 |    2.269 |   959.13 |
  |  4096 |    128 |    1 |   4224 |    2.235 |  1832.54 |    1.326 |    96.52 |    3.561 |  1186.06 |
  |  8192 |    128 |    1 |   8320 |    5.845 |  1401.44 |    1.352 |    94.70 |    7.197 |  1156.03 |
  | 16384 |    128 |    1 |  16512 |   17.096 |   958.38 |    1.417 |    90.33 |   18.513 |   891.94 |
  | 32768 |    128 |    1 |  32896 |   56.013 |   585.00 |    1.530 |    83.66 |   57.543 |   571.67 |

  Carnice-Qwen3.6-MoE-35B-A3B-Q4_K_S
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.205 |  2499.78 |    1.483 |    86.31 |    1.688 |   379.16 |
  |  1024 |    128 |    1 |   1152 |    0.434 |  2361.36 |    1.448 |    88.40 |    1.882 |   612.25 |
  |  2048 |    128 |    1 |   2176 |    0.947 |  2161.87 |    1.478 |    86.62 |    2.425 |   897.27 |
  |  4096 |    128 |    1 |   4224 |    2.259 |  1813.00 |    1.472 |    86.94 |    3.732 |  1131.98 |
  |  8192 |    128 |    1 |   8320 |    5.892 |  1390.42 |    1.505 |    85.06 |    7.397 |  1124.85 |
  | 16384 |    128 |    1 |  16512 |   17.397 |   941.77 |    1.568 |    81.61 |   18.965 |   870.63 |
  | 32768 |    128 |    1 |  32896 |   56.296 |   582.07 |    1.690 |    75.74 |   57.986 |   567.31 |

  Nemotron-Cascade-2-30B-A3B-IQ4_XS
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.195 |  2622.33 |    0.972 |   131.69 |    1.167 |   548.30 |
  |  1024 |    128 |    1 |   1152 |    0.407 |  2514.76 |    0.934 |   137.10 |    1.341 |   859.16 |
  |  2048 |    128 |    1 |   2176 |    0.854 |  2396.99 |    0.942 |   135.90 |    1.796 |  1211.42 |
  |  4096 |    128 |    1 |   4224 |    1.895 |  2161.89 |    0.953 |   134.36 |    2.847 |  1483.50 |
  |  8192 |    128 |    1 |   8320 |    4.593 |  1783.70 |    0.967 |   132.43 |    5.559 |  1496.60 |
  | 16384 |    128 |    1 |  16512 |   12.213 |  1341.53 |    0.996 |   128.56 |   13.209 |  1250.10 |
  | 32768 |    128 |    1 |  32896 |   36.998 |   885.66 |    1.059 |   120.89 |   38.057 |   864.39 |

  On mbp M2 Max

  Qwen3.6-35B-A3B-UD-Q6_K_XL
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.540 |   947.31 |    2.489 |    51.42 |    3.030 |   211.22 |
  |  1024 |    128 |    1 |   1152 |    0.951 |  1077.21 |    3.237 |    39.54 |    4.188 |   275.10 |
  |  2048 |    128 |    1 |   2176 |    2.994 |   684.10 |    3.139 |    40.77 |    6.133 |   354.80 |
  |  4096 |    128 |    1 |   4224 |    6.245 |   655.85 |    3.210 |    39.88 |    9.455 |   446.75 |
  |  8192 |    128 |    1 |   8320 |   12.411 |   660.08 |    3.284 |    38.98 |   15.694 |   530.13 |
  | 16384 |    128 |    1 |  16512 |   28.321 |   578.51 |    3.584 |    35.71 |   31.905 |   517.53 |
  | 32768 |    128 |    1 |  32896 |   65.725 |   498.56 |    4.029 |    31.77 |   69.754 |   471.60 |

  Nemotron-Cascade-2-30B-A3B-Q8_0
  |    PP |     TG |    B |   N_KV |   T_PP s | S_PP t/s |   T_TG s | S_TG t/s |      T s |    S t/s |
  |-------|--------|------|--------|----------|----------|----------|----------|----------|----------|
  |   512 |    128 |    1 |    640 |    0.528 |   969.13 |    2.036 |    62.87 |    2.564 |   249.59 |
  |  1024 |    128 |    1 |   1152 |    1.079 |   948.84 |    3.201 |    39.99 |    4.280 |   269.15 |
  |  2048 |    128 |    1 |   2176 |    3.390 |   604.10 |    2.952 |    43.36 |    6.342 |   343.11 |
  |  4096 |    128 |    1 |   4224 |    6.756 |   606.28 |    2.991 |    42.79 |    9.747 |   433.35 |
  |  8192 |    128 |    1 |   8320 |   13.647 |   600.30 |    3.061 |    41.81 |   16.708 |   497.97 |
  | 16384 |    128 |    1 |  16512 |   29.491 |   555.56 |    3.414 |    37.50 |   32.905 |   501.81 |
  | 32768 |    128 |    1 |  32896 |   65.867 |   497.49 |    3.663 |    34.95 |   69.530 |   473.12 |

Glad to see other people using it. Saved my life, was going crazy click-clicking to nab the right window. Now Cmd-1..9 brings to focus a window of my chosen application. (Chrome) In case it helps someone else, myself and Codex iterating over time https://github.com/ljubomirj/dotfiles/blob/main/.hammerspoon.... Cmd-1..9 switches over focuses to a particular window, Cmd-0 presents an (ugly; but suffices) dialog box to select the window with arrows (of the App of interest - Chrome for me atm) to switch to. But more important - to see what window what Window name is recalled by the particular Cmd-1..9 shortcut. Option-arrows shuffle window-to-key ordering. I right-click-Name Window my windows. Think back now - on restart they may even be preserved?? Don't recall re/naming them manually recently. (possible I've forgotten though)

Say more please if you can. How/why is ik_llama.cpp faster then mainline, for the 27B dense? I'd like to be able to run 27B dense faster on a 24GB vram gpu, and also on an M2 max.

Everyone should do the calculation for themselves. I too pay for couple of subs. But I'm noticing having an agent work for me 24/7 changes the calculation somewhat. Often not taken into account: the price of input tokens. To produce 1K of code for me, the agent may need to churn through 1M of tokens of codebase. IDK if that will be cached by the API provider or not, but that makes x5-7 times price difference. OK discussion today about that and more https://x.com/alexocheema/status/2020626466522685499

https://ljubomirj.github.io small personal ~/public_html

+1. For every one like the author of the blog post, it's likely to be another one in the opposite direction. But they will be unlikely to write a post about that. I too found weighting 'spend time with human persons v.s. with my own thoughts, or programming and writing, or reading a paper or a post, or listening to a podcast while walking in nature' lately come down on the side away from humans. So far - it's been way more interesting. When/if that changes and becomes boring - will think what next and change.

I imagine - and sure hope so - everyone trains on everything else. Distillation - ofc if one has bigger/other models providing true posterior token probabilities in the (0,1) interval (a number between 0 and 1), rather than 1-hot-N targets that are '0 for 200K-sans-this-token, and 1 for the desired output token' - one should use the former instead of the latter. It's amazing how as a simple as straightforward idea should face so much resistance (paper rejected) and from the supposedly most open minded and devoted to knowing (academia) and on the wrong grounds ('will have no impact on industry'; in fact - it's had tremendous impact on industry; better rejection wd have been 'duh it is obvious'). We are not trying to torture the model and the gpu cluster to be learning from 0 - when knowledge is already available. :-)

Nah - given the ergonomics + economics, local coding models are not atm that viable. I like all things local even if just for safety of keeping healthy competitive ecosystem. And I can imagine really specialised uses cases where I run an 8B not-so-smart model to process oodles of data on my local 7900xtx or similar. Got older m2 mbp with 96gb (v)ram and try all things local that fit. Usually LMStudio for the speed add in MLX format models on ASI (as end point; plus chat for vibes test; LMStudio omission from the OP blog post makes me question the post), or llama.cpp for GGUF (llama.cpp is the OG; excellent and universal engine and format; recently got even better). Looking at how agents work - an agent smarts of Claude Code or Codex in using the tools feels like it's half its success (the other half the underlying LLM smarts). From the training on baked in 'Tool Use & Interleaved Thinking' on the right tools in a right way, to the trivial 'DONOTDO bad idea to fill your 100K useful context with random content of multi-MB file as prompt'. The $20/mo plans are insanely competitive. OpenaI is generous with Codex, and in addition to terminal that I mostly use, there is the VSCode addon as well as use in Cline or Roo. Cursor offers in-house model fast and good, insane economy reading large codebases, as well BYOK to latest-greatest LLMs afaik. Claude Code $20/mo is stingy with quotas, but can be supplement with Z.ai standing in - glm-4.7 as of yesterday (saw no difference glm-4.6 v.v. sonnet-4.5 already v.good). It's a 3 lines change to ~/.claude/settings.json to flip Z.ai-Anthropic back and forth at will (e.g. when paused on one to switch to the other). Have not tried the Cerebras high tok/s but wd love to - not waiting makes a ton of difference to productivity.

Not for SSD specifically, but I assume the compact design doesn't hurt: duckdb saved my sanity recently. Single file, columnar, with builtin compression I presume (given in columnar even simplest compression maybe very effective), and with $ duckdb -ui /path/to/data/base.duckdb opening a notebook in browser. Didn't find a single thing to dislike about duckdb - as a single user. To top it off - afaik can be zero-copy 'overlayed' on the top of a bunch of parquet binary files to provide sql over them?? (didn't try it; wd be amazing if it works well)

So far so good - and I say this as one voting remain. The only gripe I have is that our domestic doomers were even more stupid than the EU ones. Ours were the progenitors of many of EU dumb ideas. So even outside EU, we in the UK not only did not repeal the utterly imbecilic laws we inherited. No - we added even more stupid laws. Consequence being people are put in jail for writing stuff on the Internet. I hope someone puts in jail the lawmakers that voted for these laws. To the cheering of and with public support, it must be said. It was not without consent, it was not only bi-party, but omni-party consent.

Yeah - IDK why it never makes it to these lists. R too. Matlab being 2D matrix first/default gets it right for me there. IK matrices trivially translate to arrays, still: find 2D to be extra expressive on human level, for zero price paid. I get it it's all the same to the cpu. 2D rows-columns rectangle of data being the simplest data structure both necessary and sufficient covering a 1) matrix 2) spreadsheet 3) SQL table 4) directed graph of nodes and edges. (in the past I've read someplace that lists are for pie eaters, but wouldn't know myself

Yes - so telling hard truths is not for the benefit of the listener, it's for the benefit of the speaker mostly. That's a major point: if I see, but I don't tell, if I have private truths and public lies, it's one small victory for untruths. However much I think I'm not - I'm co-opted in the big lie machine. There is quite a lot of experience with this acquired during the totalitarian communist regimes that existed in eastern Europe <1990s. And a minor point is: the listener may switch off, but a minuscule part of the message may make it's way. May implant a tiny seed of doubt, admittedly very very unlikely. But it's not totally futile. Even if the speaker may decide the price to be paid is too high, for too little gain. (lots of the time)

Iceland - sorry for the confusion. (poor spelling)

Eh - for the world's audiences: EU is not Europe. I geddit how/why why people equate eu == europe - it would simplify things for all, one niggle less to consider. But - it ain't so, for better or worse. There are countries in Europe, that can't be members of the European Union, or could be, but don't want to be members. (e.g. UK, probably Island, Switzerland, some of the Nordics) There are no countries in the European Union, that are not part of Europe. So EU <= Europe. (unsurprisingly)

I think you are correct there - the majority of the public don't care. They just try to get about doing their daily business and act the best they can under circumstances. So we just click "Accept" to any popup banner make it go away, accept "All cookies" 100 times every day, use Google mail/map/photos/drive and that all involves giving away data, even if in words we say we don't want to give data. So yes obviously the public by necessity act in a rational way, doing cost-benefit analysis. While a cadre of privacy obsessives have made my life worse by lobbying and having their bad ideas codified in the UL laws. Wrote about my experience in the UK medical systems here https://news.ycombinator.com/item?id=45066321