Application map · explore the portfolio
One portfolio. Focused applications.
36 applications, from AI foundations and agent systems to digital twins and humanoid robotics. Explore the parent applications and their labs below. The three-letter code is the permanent key between each application, repository, and aserdargun.com subdomain.
| Code | Application | Repository | Address |
|---|---|---|---|
aia | AI Ecosystem AtlasAn evidence-backed research console for comparing AI products and developer ecosystems.Current
Evidence & related knowledge | aia-aserdargun-com opens in a new tab | aia.aserdargun.com opens in a new tab |
cpu | CPU Atlas — RISC-V Processor ArchitectureA visual knowledge and learning guide to how processors work. Twelve lessons connect RISC-V instructions, registers, datapaths, pipelines and caches with logic gates, transistors and physical layouts. Interactive RTL/WASM examples help you explore the concepts step by step.Review soon
Evidence & related knowledgeGuiding questionHow do processors work, from instructions to transistors? | cpu-aserdargun-com opens in a new tab | cpu.aserdargun.com opens in a new tab |
rtx | RTX Spark Atlas — N1X 128 GBExplore Arm CPU, Blackwell GPU, NVLink-C2C and 128 GB unified memory in a bilingual Blender and Three.js atlas. Select components, separate layers and animate data flows. A representative educational model based on NVIDIA documentation, not a measured physical blueprint.Current
Evidence & related knowledgeGuiding questionHow do Arm and CUDA share work and memory in N1X? | rtx-aserdargun-com opens in a new tab | rtx.aserdargun.com opens in a new tab |
llm | LLM Runtime & Serving AtlasA source-backed field guide to LLM runtimes and serving engines across seven architectural layers, with an interactive learning layer.Current
Evidence & related knowledge | llm-aserdargun-com opens in a new tab | llm.aserdargun.com opens in a new tab |
tfl | Sub-application of LLMTFL — Token Flow LaboratoryExplore nine deterministic serving experiments and ten guided checkpoints: queueing, prefill, KV reservations, decode and token delivery. Compare scheduling, latency and throughput; token text is scripted and timings are synthetic, with no model execution or hardware benchmark.Current
Evidence & related knowledgeGuiding questionHow do finite serving resources shape a request’s path to its next token? | tfl-aserdargun-com opens in a new tab | tfl.aserdargun.com opens in a new tab |
hns | Harness Engineering ObservatoryA bilingual, source-backed observatory comparing seven agent-system layers: execution, tools, context, lifecycle, observability, verification, and governance. Evidence and editorial interpretation remain separate; no universal ranking.Current
Evidence & related knowledge | hns-aserdargun-com opens in a new tab | hns.aserdargun.com opens in a new tab |
dpl | Sub-application of HNSDPL — Decision Plane LaboratoryAn independent deterministic simulation for the decision plane: take a short path, evaluate further, ask for evidence, wait for approval, abstain or block an action.Current
Evidence & related knowledgeGuiding questionWhen should an agent decide, ask, wait, or stop? | dpl-aserdargun-com opens in a new tab | dpl.aserdargun.com opens in a new tab |
cul | Sub-application of HNSCUL — Computer Use LaboratoryObserve, act, verify: synthetic computer-use scenarios that keep targeting, approval and result checking inside one controlled browser environment.Current
Evidence & related knowledgeGuiding questionHow does an agent act on a screen, and how is the result verified? | cul-aserdargun-com opens in a new tab | cul.aserdargun.com opens in a new tab |
aos | Sub-application of HNSAgent Operating SystemA public guide to a target agent operating-system architecture: local models, an isolated computer, an open state and human control.Current
Evidence & related knowledgeGuiding questionWhat would an agent operating system need to stay observable and human-controlled? | aos-aserdargun-com opens in a new tab | aos.aserdargun.com opens in a new tab |
arl | Sub-application of HNSARL — Agent Runtime LaboratoryInspect a deterministic agent run through harness, context, security and evaluation lenses. Explore tool use, evidence and one-use approval; all actions remain inside the simulation.Current
Evidence & related knowledgeGuiding questionWhat happens between intent and action? | arl-aserdargun-com opens in a new tab | arl.aserdargun.com opens in a new tab |
agr | Sub-application of HNSAgoraA deliberation panel of OpenRouter’s free models. Every seat answers the proposition alone, then answers the other seats by name, and the dissenters are invited back once more. Agora counts the discussion itself and reports whether the panel converged, measured against a bar it declares in advance. Agreement is reported as agreement, never as correctness.Current
Evidence & related knowledgeGuiding questionDo free models converge on anything, or only agree when they were never tested against each other? | agr-aserdargun-com opens in a new tab | agr.aserdargun.com opens in a new tab |
sec | AI Systems Security ObservatoryA bilingual, evidence-aware observatory for tracing AI-agent trust from model intent through identity, authorization, constrained action, audit, and incident recovery.Current
Evidence & related knowledge | sec-aserdargun-com opens in a new tab | sec.aserdargun.com opens in a new tab |
ctx | Context & Knowledge EngineeringA bilingual, source-backed field guide to the information system that runs before a model answers, from ingestion and retrieval through citation, caching, and memory.Current
Evidence & related knowledge | ctx-aserdargun-com opens in a new tab | ctx.aserdargun.com opens in a new tab |
mem | Sub-application of CTXMEM — Agent Memory LaboratoryExplore agent memory across six synthetic scenarios and three memory policies: recall, source and version links, correction, expiry, deletion and JSON export.Current
Evidence & related knowledgeGuiding questionWhat should an agent remember, and when should it recheck? | mem-aserdargun-com opens in a new tab | mem.aserdargun.com opens in a new tab |
evl | AI Evaluation & Reliability LabA bilingual, evidence-aware workbench for evaluation contracts and deterministic AI release decisions across output, trajectory, outcome, robustness, safety, and operations.Current
Evidence & related knowledge | evl-aserdargun-com opens in a new tab | evl.aserdargun.com opens in a new tab |
usl | Unsloth Studio Learning AtlasA bilingual learning atlas with eight lessons, a 12-week roadmap, teaching labs and historical experiment evidence for LoRA/QLoRA, datasets, evaluation and local deployment. No live GPU or training; ADP extends adaptation practice with synthetic experiments.Current
Evidence & related knowledge | usl-aserdargun-com opens in a new tab | usl.aserdargun.com opens in a new tab |
adp | Sub-application of USLADP — Model Adaptation LaboratoryExplore Full FT, LoRA and QLoRA through parameter counts, memory estimates, data preparation and synthetic training/evaluation experiments. USL provides the adaptation foundation; an optional DCL handoff carries educational workload metadata into deployment planning. No real model is trained or deployed.Current
Evidence & related knowledgeGuiding questionHow does a base model become adapted to a task? | adp-aserdargun-com opens in a new tab | adp.aserdargun.com opens in a new tab |
gpu | GPU Kernel Engineering — Kernel AtlasA bilingual 12-week interactive atlas for CUDA, Triton, GPU memory, LLM operators, correctness, profiling, inference, and multi-GPU systems.Current
Evidence & related knowledge | gpu-aserdargun-com opens in a new tab | gpu.aserdargun.com opens in a new tab |
pol | Sub-application of GPUPOL — Programming LanguagesEnglish-only comparisons of the same tasks across programming languages: types, memory, paradigms and concurrency. Programming foundations that support GPU learning, but are not limited to GPU. Executed examples and hand-reviewed examples remain distinct.Current
Evidence & related knowledgeGuiding questionHow do languages express the same program differently? | pol-aserdargun-com opens in a new tab | pol.aserdargun.com opens in a new tab |
gex | Sub-application of GPUGPU Execution ExplorerTrace GPU anatomy, kernel launches, warp masks, memory access, and matrix tiles through six interactive lessons. Results come from an educational execution model, not profiler measurements.Current
Evidence & related knowledgeGuiding questionHow does a kernel become hardware execution? | gex-aserdargun-com opens in a new tab | gex.aserdargun.com opens in a new tab |
cld | Cloud Provider Cost ComparisonA source-backed decision tool comparing pre-tax USD costs for cloud services available from Türkiye.Current
Evidence & related knowledge | cld-aserdargun-com opens in a new tab | cld.aserdargun.com opens in a new tab |
lcl | Local Compute LabA bilingual, source-backed workbench matching open-weight model requirements to NVIDIA, AMD, and Apple devices. Explore dated market observations, memory assumptions, compatibility evidence, and concept cards; compare local and cloud deployment assumptions in the shared DCL lab.Current
Evidence & related knowledge | lcl-aserdargun-com opens in a new tab | lcl.aserdargun.com opens in a new tab |
dcl | Shared laboratory of CLD + LCLDeployment Choice LaboratoryThe shared laboratory of CLD and LCL. Compare local hardware, cloud GPUs, managed inference and token APIs against workload, memory, privacy and cost assumptions. Hard constraints and educational estimates explain conditional choices; no live pricing or universal winner is claimed.Current
Evidence & related knowledgeGuiding questionWhere should this workload run, and under which assumptions? | dcl-aserdargun-com opens in a new tab | dcl.aserdargun.com opens in a new tab |
wfm | World Models AtlasA living research atlas tracing how world models connect perception, prediction, planning, and action through primary sources.Current
Evidence & related knowledge | wfm-aserdargun-com opens in a new tab | wfm.aserdargun.com opens in a new tab |
wml | Sub-application of WFMWorld Model LaboratoryExplore observation, belief, prediction, planning, and counterfactual branches in WFM’s experiment laboratory. Compare three hand-coded predictors against a simulated world; these are not trained foundation models.Current
Evidence & related knowledgeGuiding questionHow does a prediction change when beliefs or actions change? | wml-aserdargun-com opens in a new tab | wml.aserdargun.com opens in a new tab |
swi | SWI — Swarm IntelligenceA bilingual swarm-intelligence atlas with eight biology dossiers, 21 selected studies and eight static agent experiment templates. SWI observes and explains; ANT and BEE own the colony simulations. Recipes export Markdown/JSON without running agents; the agenda stays in the browser.Current
Evidence & related knowledge | swi-aserdargun-com opens in a new tab | swi.aserdargun.com opens in a new tab |
ant | Sub-application of SWIANT — Ant Colony Intelligence LaboratorySWI’s companion ant-colony lab: one reproducible experiment, three guided exercises and run replay. Explore local rules, food-pheromone trails and evaporation. Home chemical is observational; returning ants use idealized path integration. No species calibration or guaranteed shortest path; synchronized A/B is future work.Current
Evidence & related knowledgeGuiding questionHow can local chemical traces coordinate foraging without a central plan? | ant-aserdargun-com opens in a new tab | ant.aserdargun.com opens in a new tab |
bee | Sub-application of SWIBEE — Honey Bee Collective Intelligence LaboratoryFour deterministic, bilingual experiments within SWI’s research family: explore local discovery, private memory, dance recruitment and resource allocation with a same-seed dance-off control, guided lessons and JSON replay. An Apis mellifera-inspired educational model; results use model units, not field measurements or biological predictions.Current
Evidence & related knowledgeGuiding questionHow do private memory and local dance signals shape collective foraging in this model? | bee-aserdargun-com opens in a new tab | bee.aserdargun.com opens in a new tab |
itl | Industrial Twin LabAn English research publication with authored deterministic P-101 evidence fixtures. Explore pump anatomy in PDT and computed synthetic decision experiments in DTR; each application has separate models and state, with no field connection or control authority.Current
Evidence & related knowledge | itl-aserdargun-com opens in a new tab | itl.aserdargun.com opens in a new tab |
dtr | Sub-application of ITLDTR — Digital Triplet LaboratoryExplore six synthetic pumping scenarios with an independent twin, three 60-second alternatives, explicit human approval, prediction–outcome comparison and JSON experiment records. Runs entirely in the browser with explainable rules; no LLM, field telemetry or machine commands.Current
Evidence & related knowledgeGuiding questionWhen does a twin prediction support a decision, and what does the outcome reveal? | dtr-aserdargun-com opens in a new tab | dtr.aserdargun.com opens in a new tab |
pdt | Sub-application of ITLP-101 Interactive Digital TwinExplore P-101 through four views, nine components, eight sensor locations and five authored pump conditions. Compare qualitative synthetic signals; this ITL teaching companion has no live telemetry or shared experiment state.Current
Evidence & related knowledgeGuiding questionHow do pump components relate to operating conditions? | pdt-aserdargun-com opens in a new tab | pdt.aserdargun.com opens in a new tab |
eng | Open Humanoid EngineeringThe learning system’s humanoid engineering horizon: an English manifesto following ITL, with HEX as its exploration companion. The proposed four-year spine, AI lab pipeline, and dated 2026 frontier signals are primary sources of truth. Seven specializations lead to portfolios; no degree, running AI laboratory, or completed physical robot is claimed.Current
Evidence & related knowledgeGuiding questionHow can an AI-native team turn a digital humanoid design into evidence for human review? | eng-aserdargun-com opens in a new tab | eng.aserdargun.com opens in a new tab |
hex | Sub-application of ENGHumanoid Engineering ExplorerExplore humanoid structure, joints, actuation, sensors, and power/data paths through ENG’s interactive 3D learning object. Movements are kinematic teaching aids, with no asserted dynamics or hardware performance.Current
Evidence & related knowledgeGuiding questionHow do body, joints, sensing, and control work together? | hex-aserdargun-com opens in a new tab | hex.aserdargun.com opens in a new tab |
vis | Computer Vision Knowledge BankA source-backed computer-vision field guide across seven perception layers, each layer naming what it is not for, plus a deterministic laboratory that measures the same ideas against a per-pixel answer key. The knowledge bank explains and cites; the laboratory measures. Classical operators, a CNN trained at runtime, monocular depth cues and optical flow all run on a CPU path by default and an honestly probed WebGPU path when available.Current
Evidence & related knowledgeGuiding questionWhat does a camera actually know, and how much of that survives contact with the real world? | vis-aserdargun-com opens in a new tab | vis.aserdargun.com opens in a new tab |
cvl | Sub-application of VISCVL — Computer Vision LaboratoryA deterministic laboratory that measures eight computer-vision layers against a per-pixel answer key painted in the same pass as the synthetic scene. Classical operators, a convolutional network trained at run time, monocular depth cues and sparse optical flow run on an f64 CPU path by default and on an honestly probed WebGPU path when an adapter exists. The VIS knowledge bank explains; CVL measures.Current
Evidence & related knowledgeGuiding questionWhich failure does this operator reach first, and can it be measured before it ever meets a real camera? | cvl-aserdargun-com opens in a new tab | cvl.aserdargun.com opens in a new tab |
ilm | ILMEK — A World RewovenA third-person exploration game that turns this learning system into playable ground: every registered application is a region with its own puzzle mechanic, and the world is restored by rediscovering how systems relate and reconnecting them. Geometry, texture and sound are generated procedurally at runtime. There are no accounts, API keys, paid services or live model calls, and the in-game puzzles are teaching models rather than the applications' own measurements.Current
Evidence & related knowledgeGuiding questionCan this whole system be learned by playing it instead of reading it? | ilm-aserdargun-com opens in a new tab | ilm.aserdargun.com opens in a new tab |
No applications match. Try another term or choose all layers.