Steve Leggett
@cdnsteve 329 X followers cdnsteveroboticforce.io First product launched 9mo ago
- Public revenue
- Undisclosed
- No public revenue across products
Products
| Product | Visits · Oct | Actions |
|---|---|---|
| WorkSignalThe application is a form. The screen is a conversation. | <5K | |
| TokensteadRun local AI models on hardware you own | <5K | |
| YouBrokeProdOnline training platform for incident response training. Engineers | <5K | |
| LearningToLearning to pass the interview | <5K |
Timeline
Nov 17 · 4d ago
Tokenstead reports GLM-5.3-Flash benchmarks on four DGX SparksReferrals & Backlinks
The report tests Z.ai's GLM-5.3-Flash using Bizuayeu's XSFP4 recipe and TensorFold 1.2.9 on four ring-connected DGX Sparks. It reports 86.1 tok/s in a temperature-zero decode test and 52.0 tok/s in a sampled standard test.
Nov 16 · 5d ago
Tokenstead reports pruned Drama 3 benchmarks on RTX 5060 TiReferrals & Backlinks
Tokenstead reports a real-time factor of 0.383 on 24 texts for plotarmordev's pruned, 3-bit Drama 3 running on an RTX 5060 Ti. The report records peak device memory of 14.77 GiB and notes that differences in codecs, reference voices, quantization and engines prevent a pure hardware comparison.
Tokenstead publishes TontaubeV1.1 and ZONOS3-Mini voice benchmarksReferrals & Backlinks
Tokenstead compares TontaubeV1.1 and ZONOS3-Mini on an RTX 5090 using five sentences. It reports audible speech after 470 ms for TontaubeV1.1 and 254 ms for ZONOS3-Mini, with generation speeds of 15.6x and 36.6x real time respectively.
Tokenstead reports Qwen3.8 Flash Next benchmarks with thinking enabledReferrals & Backlinks
The report shows MMLU-Pro accuracy rising from 69.5% to 88.5% and GPQA Diamond from 52.0% to 88.4% with thinking enabled. The thinking-enabled run used TensorFold 1.2.9, while the thinking-disabled run used 1.2.8 with a smaller token limit.
Tokenstead reports Edge0-145B benchmarks on a 16 GB RTX 5060 TiReferrals & Backlinks
The report lists Edge0-145B Q3_K decode performance of 12.9 tok/s on an RTX 5060 Ti 16 GB, with 554 seconds to the first token for a 32K-token prompt. Its RTX 5090 comparison lists 134.3 tok/s and 8.0 seconds to the first token.
Questions & evidence
- What products has Steve Leggett built?
- Steve Leggett tracks 4 standalone software products on EarlyLore (4 currently live), led by WorkSignal, Tokenstead, YouBrokeProd.
- How much website traffic do Steve Leggett's products get?
- Across tracked standalone domains with web traffic estimates, Steve Leggett's portfolio reaches 5.5K combined monthly visits in each site's latest observed month. Coverage excludes shared third-party hosts and store listings.
Related indie founder research guides