[{"data":1,"prerenderedAt":122},["ShallowReactive",2],{"maker-HenryNdubuaku":3},[4,36,57,81,103],{"id":5,"name":6,"line":7,"ai":8,"cat":9,"year":10,"points":11,"comments":12,"rec":13,"s":14,"top":15,"kw":16,"maker":22,"domain":23,"tier":24,"attempts":25,"strip":26,"url":27,"mix":28,"praise":33,"objection":34,"sub":35},44524544,"Cactus","Ollama for Smartphones",false,"ai",2025,231,82,"curious",0.04,7,[17,18,19,20,21],"on-device llm","mobile llm","gguf","flutter llm","react native llm","HenryNdubuaku","github.com","notable",0,"0313305353335135133130531553350113551111","https:\u002F\u002Fgithub.com\u002Fcactus-compute\u002Fcactus",[29,30,31,32],11,39,1,12,"Simple, clean API that is easy to integrate","Wraps llama.cpp, so the framework is thin and the 'Ollama for phones' framing is questioned","on-device llm framework",{"id":37,"name":38,"line":38,"ai":8,"cat":39,"year":40,"points":41,"comments":42,"rec":13,"s":43,"top":44,"kw":45,"maker":22,"domain":23,"tier":24,"attempts":25,"strip":51,"url":52,"mix":53,"praise":55,"objection":56,"sub":49},47036063,"Maths, CS and AI Compendium","learning_science",2026,88,26,0.25,18,[46,47,48,49,50],"maths textbook","intuition first","machine learning theory","open textbook","linear algebra","000503035053000000000000000","https:\u002F\u002Fgithub.com\u002FHenryNdubuaku\u002Fmaths-cs-ai-compendium",[54,54,25,25],3,"Readers find the chapters clear and helpful for refreshing maths","Spelling debate over 'Maths' versus 'Math' in the title",{"id":58,"name":59,"line":60,"ai":8,"cat":9,"year":40,"points":61,"comments":62,"rec":63,"s":64,"top":31,"kw":65,"maker":22,"domain":23,"tier":71,"attempts":25,"strip":72,"url":73,"mix":74,"praise":78,"objection":79,"sub":80},48111896,"Needle","We Distilled Gemini Tool Calling into a 26M Model",776,211,"mixed",0.19,[66,67,68,69,70],"function calling","tool use","on-device model","distillation","small language model","legend","3145033535330505430335133311403333554530","https:\u002F\u002Fgithub.com\u002Fcactus-compute\u002Fneedle",[75,76,31,77],29,45,16,"Useful for on-device uses like Home Assistant and Siri-like assistants","Failed on chained timer request and gave wrong output for some queries","function calling model",{"id":82,"name":83,"line":84,"ai":8,"cat":9,"year":40,"points":85,"comments":86,"rec":13,"s":87,"top":88,"kw":89,"maker":22,"domain":23,"tier":24,"attempts":25,"strip":94,"url":95,"mix":96,"praise":100,"objection":101,"sub":102},49010782,"Cactus Hybrid","We taught Gemma 4 to know when it's wrong",191,44,0.15,8,[90,91,92,93],"gemma 4","on-device ai","confidence score","llm calibration","1000000000000533300503030505050100053033","https:\u002F\u002Fgithub.com\u002Fcactus-compute\u002Fcactus-hybrid",[97,98,25,99],6,10,2,"Interesting idea and useful on-device approach","Calling it knowing when wrong overstates what a confidence score can show","on-device language model confidence",{"id":104,"name":105,"line":106,"ai":8,"cat":9,"year":40,"points":107,"comments":108,"rec":13,"s":109,"top":31,"kw":110,"maker":22,"domain":113,"tier":71,"attempts":99,"strip":114,"url":115,"mix":116,"praise":119,"objection":120,"sub":121},49246804,"Needle2","14MB agentic LLM for phones, wearables, smart home and robots",537,185,-0.07,[111,17,112,66,70],"tool calling model","edge ai","cactuscompute.com","3353152131431131315532142133335332334513","https:\u002F\u002Fcactuscompute.com\u002Fneedle",[44,117,118,75],57,4,"Impressively small model that runs locally in WASM","Gives wrong tool calls for simple commands like thermostat and lights","small on-device tool-calling model",1793991306471]