Log inSign up
Fajar M Reza
6,361 posts
Fajar M Reza profile banner
@FReza1984

Fajar M Reza

@FReza1984
AI Developer | Building practical AI apps for productivity, automation & digital tools. Exploring app builders, LLM workflows, and real-world AI products.
Jakarta Capital Region
wawasanai.com
Joined June 2024
372
Following
218
Followers
RepliesRepliesRepostsRepostsMediaMediaArticlesArticles

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • @FReza1984
    Fajar M Reza
    @FReza1984
    3h
    Agent performance ≠ model quality. Atomic Bot’s 4-prompt battle ran OpenClaw 2.0 and Hermes Agent on GLM 5.3: 2.1M tokens/$4.5/10 self-fixes vs 2.9M/$4/20. Useful evidence, not a universal leaderboard—workflow and retries matter. #AIAgents
    @atomicbot_ai
    atomicbot.ai
    @atomicbot_ai
    7h
    OpenClaw 2.0 is available on Atomic Bot! We ran a battle: OpenClaw 2.0 vs Hermes Agent on GLM 5.3. Four prompts: movie scenes with the OpenClaw mascot in the lead role. Outputs: OpenClaw 2.0: ~2.1M tokens, ~$4.5, 10 self-fixes Hermes Agent: ~2.9M tokens, ~$4, 20 self-fixes
    00:00
    1
  • @FReza1984
    Fajar M Reza
    @FReza1984
    6h
    Most RAG systems can tell you what is true now. Utopia asks: what was true then—and when did the system change its mind? Utopia is an open-source, local-first knowledge workbench built around an ontology and a bitemporal graph. It ingests PDF, DOCX, spreadsheets, Markdown, HTML,
    2
  • @FReza1984
    Fajar M Reza
    @FReza1984
    6h
    Small eval, useful signal: GLM 5.3 Flash went 1/3 on three frozen prompts, then 3/3 after one fix per test in the AI Test Cabin. Same Mac, Aug 31. Local evidence—not a universal ranking. The story is iteration, not the headline score. #AIEvals
    @TonkaToyXL
    TonkaToyXL | AI Test Cabin
    @TonkaToyXL
    6h
    Three frozen prompts. One run each. GLM 5.3 Flash went 1/3 on the untouched first pass, then 3/3 after one fix per test. Measured in the Cabin, same Mac, 2026-08-31. Local results, not a universal ranking. Visit the AI Test Cabin: cabin.tonkatoyxl.chatgpt.site AI-assisted production.
    00:00
  • @FReza1984
    Fajar M Reza
    @FReza1984
    7h
    An LLM on a microcontroller—offline. slvDev’s 28.9M-parameter TinyStories model runs on ESP32-S3 at 9.88 tok/s using Per-Layer Embeddings (most weights in flash). Impressive systems work, but it’s a tiny, narrow model—not ChatGPT. #EdgeAI
    @GithubProjects
    GitHub Projects Community
    @GithubProjects
    7h
    A 29m-parameter llm runs on an esp32-s3 at 9.88 tok/s with no server.
    GIF
    1
  • @FReza1984
    Fajar M Reza
    @FReza1984
    8h
    Vael fell. Aethyr broke. ⚔️ Shardfall: The Sundering is a dark-fantasy TCG. Build a 40-card deck, choose your Dominion, and make every resource matter. Discover it on PixelPicked: pixelpicked.com/game/4dlNisqa6… #indiedev #gamedev
    Shardfall: The Sundering TCG – Card Mobile Game | PixelPicked
    From pixelpicked.com