MCP/Tool-Usage with PALM - Pascal-native LLM interference engine
youtu.be/lml0V0zooLM
#mcp #ai #llm #objectpascal #delphi #fpc #freepascal
@rosseaux.bsky.social
Programming for food, indie game programmer, hobby musician, demoscener under the handle BeRo in the demogroup Farbrausch.
MCP/Tool-Usage with PALM - Pascal-native LLM interference engine
youtu.be/lml0V0zooLM
#mcp #ai #llm #objectpascal #delphi #fpc #freepascal
youtu.be/K6HY_vxY6a4
PALM - Pascal-native LLM interference engine #ai #llm #pascal #lazarus #delphi #objectpascal Uses "no" llama.cpp or something like that!
A pity, as I used to hold the one well known person in high regard, which makes it all the more disappointing now. It was about whether LLMs can count letters. I provided a working demo and peer-reviewed research showing itโs possible under the right setup, but the exchange ended before it began.
09.08.2025 06:37 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0Sometimes, even in technical discussions, you can present clear evidence, working examples, and credible sources, but before any real back-and-forth can even start, the only reply is a block. And I was ready to talk openly and willing to find common ground. ...
09.08.2025 06:37 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0And read tip: "Can ChatGPT Learn to Count Letters?" Mar. 2025, pp. 96-99, vol. 58 DOI Bookmark: 10.1109/MC.2024.3488313
09.08.2025 05:40 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0... and with the right tokenizer/configuration it can effectively reconstruct letters before reasoning over them. So, while the standard LLM experience might fail on this task, itโs not an inherent impossibility of LLMs themselves. Itโs an implementation detail.
09.08.2025 05:28 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0... I can feed an unknown fantasy word, have it break it down into single letters, and count them correctly, as shown in the previous post. This works because the model has learned statistical associations between tokens and their underlying character sequences, ...
09.08.2025 05:28 โ ๐ 0 ๐ 0 ๐ฌ 1 ๐ 0Sean, you are correct that most token-based LLMs donโt literally "see" letters and operate on subword tokens, but this is not a hard limitation of all LLM setups. With the Qwen3 30B A3B model running in my own WIP inference engine (Object Pascal + x86-64 inline assembler), ...
09.08.2025 05:28 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0At least the Qwen3 30B A3B model can break an unknown fantasy word down from a set of tokens into single letters and count them correctly, as shown in the attached screenshot (running in my own WIP LLM inferencing engine, implemented in Object Pascal and x86-64 inline assembler).
09.08.2025 05:12 โ ๐ 1 ๐ 0 ๐ฌ 2 ๐ 0A short preview demonstration of PALM with llama3.2 1TB as base model - A Object-Pascal-native LLM engine with inline assembler AVX2 SIMD optimizations, running on the CPU.
www.youtube.com/watch?v=LnKC...
#llm #objectpascal #pascal #freepascal #delphi #llama #ai
@golemde.bsky.social Mein Golem Plus/Pur-Abo wurde am 05. Februar gekรผndigt, trotzdem erfolgen bis jetzt Juni weiterhin Abbuchungen. Auf E-Mails keine Reaktion, keine Rรผckzahlung. Bitte um dringende Klรคrung, inklusive Verzugszinsen.
12.06.2025 00:10 โ ๐ 0 ๐ 0 ๐ฌ 0 ๐ 0๐ฎI released a small indie puzzle game called Grid Quest.
A minimalist 8ร8 vanished-tile-platform puzzle game with procedural levels, relaxing music, and pure shader-powered visuals.
Get it on bero1985.itch.io/grid-quest
#indiegames #puzzlegame #shader #gamedev #itchio
Everyone hopping onto Assassins Creed right now take note of the options that go from "Low, Medium, High, Very High, Ultra High". Please do not shove things to "Ultra High" and complain about game perf. As in many games, "Medium" or "Low" can be amazing looking and it is merely a naming scheme.
19.03.2025 15:09 โ ๐ 246 ๐ 16 ๐ฌ 26 ๐ 5๐ง Spotify: open.spotify.com/intl-de/albu...
๐ Apple Music: music.apple.com/de/album/end...
๐ป Deezer: www.deezer.com/en/album/711...
๐ Amazon: www.amazon.de/dp/B0DWT1F4C6
๐ Tidal: tidal.com/browse/track...
๐ต YouTube Music: www.youtube.com/watch?v=9FOI...
Mein neuer Track โEndlich Wieder Frรผhlingโ zusammen mit Jon Becker als Sรคnger ist da โ diesmal im Rock-Pop-Stil statt EDM.
Hรถrt rein:
โถ๏ธ YouTube: www.youtube.com/watch?v=JIzH...
๐ง Spotify: open.spotify.com/intl-de/albu...
#Frรผhling #NeueMusik #RockPop #EndlichWiederFrรผhling
A personal message, every contribution counts: Supporting my creative journey through tough times
www.youtube.com/watch?v=oDza...
Here a quick tinymembench benchmark performance comparison between the PasRISCV (Object-Pascal, FPC 3.3.1) and RVVM (C, GCC 14.2.1) RISC-V emulators.
15.01.2025 16:43 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0Regarding how the instruction interpreter part in PasRISCV roughly works:
13.01.2025 04:07 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0RV64GC RISC-V emulator PasRISCV now with Sound and host directory mounting per 9p protocol
www.youtube.com/watch?v=Febv...
The beginning of network support in my Pascal 64-bit RISC-V PasRISCV emulator. ๐ Although I still have a few routing issues here at the moment.
08.01.2025 06:17 โ ๐ 1 ๐ 0 ๐ฌ 1 ๐ 0A small comparison of Linux boot times with PasRISCV and JIT disabled RVVM (git 56f3510be) to show that my RV64-RISCV CPU emulation in Pascal (compiled with FPC) is faster than the one in C at RVVM (compiled with Clang).
youtu.be/6h1VPY9jraM
PasRISCV demonstration of the new state, an RV64GC RISC-V emulator implemented in Object Pascal.
Now with PCIe, NVMe, PS/2 Keyboard&Mouse, VirtIO Keyboard&Mouse, VirtIO Block and FrameBuffer emulations. VirtIO GPU, VirtIO NET and VirtIO Sound are also coming.
youtu.be/rWebo5AZVLY
Good news! I've launched .masterknecht, a new free browser-based PWA tool for mastering audio using WebAssembly! Master your tracks with reference files. No ML, just efficient algorithms for fast mastering on your device. Try it & donate! masterknecht.klangknecht.com #audiomastering #PWA
06.12.2024 00:27 โ ๐ 2 ๐ 0 ๐ฌ 0 ๐ 0PasVulkan now includes an external HTML5/JS-based tool for visualizing Vulkan memory allocations. Note that PasVulkan's custom Vulkan memory manager predates AMD's first public VMA release, addressing any questions about why VMA isn't used.
12.11.2024 02:17 โ ๐ 1 ๐ 0 ๐ฌ 0 ๐ 0