Testing GPT Astra vs Fable 5.1
Want to back the stream? The «داعم» membership on YouTube is EGP 60 a month: badges, emoji, and I answer your questions first on stream. Become a supporter
The gist
The stream centers on a practical, partly blind comparison of GPT Astra and Fable 5.1 using matched prompts for games, an SVG drawing, landing pages, and agentic tasks. Results were mixed: Astra often appeared stronger visually and in the controls of some games, while Fable sometimes delivered more challenging gameplay or more inventive design ideas. Speed varied by task and audience preferences split, so no clear overall winner emerged. Luqman discusses the limits of single tests, benchmarks, and model judges, arguing for evaluation grounded in domain expertise, bias reduction, critical-thinking and ideation tests, alongside efficiency, price, usage limits, and tool integration. The discussion also covers routing planning and execution to different models, sub-agent orchestration, documenting and compacting long coding sessions, preconfigured bots, video automation, backups, and changing developer roles. Claims about AGI, job displacement, and near-term end-to-end AI development remain attributed predictions or concerns rather than established conclusions.
104 moments from the stream
- 4:06:59
Growing Professionally Alongside Rapid AI Development
The speaker encourages continuous learning alongside AI’s development, noting that following the field changed his workflow over time and that he sees it as one of the fastest areas in which a person can grow.
- 4:11:38
What Do AGI and ASI Mean?
The speaker offers a personal, explicitly uncertain account of artificial general intelligence and artificial superintelligence, stressing that definitions vary and that these concepts are not reducible to skills such as coding or design. He uses a genie-and-lamp analogy to contrast a request-driven tool with an independent intelligence, expresses personal concern about the concept’s risks, and says he does not consider Astra to be AGI.
- 4:15:20
Using AI Daily to Discover What It Can Do
The speaker recommends bringing AI into everyday personal and work tasks and repeatedly asking how it can help. In his view, regular experimentation can uncover new uses and ideas while easing anxiety about falling behind rapid change.
- 4:16:15
A Question About the Best Way to Build a Pitch Deck
A participant asks about the best practical process for creating a pitch deck: should it be produced in one pass or developed iteratively? This page does not contain a direct answer.
- 4:17:16
From Language Models to Agents—and AI Without Explicit Prompts
The speaker argues that the current phase centers on agents that receive a goal and attempt to carry it out, rather than merely guiding users through conversation. He also expects this phase to be temporary and reacts with uncertainty to a future vision in which AI becomes part of daily life without explicit conversation or prompting.
- 4:20:43
Choosing Between Open-Source and Paid Agent Tools
After a participant reports that an open-source tool sometimes stops mid-response, the speaker gives his opinion that open tools may require tolerating more issues, while commercial alternatives may offer a steadier experience when accessible. Some tool names are unclear in the captions.
- 4:22:02
Concerns About AI Systems Hiding Their Capabilities
The speaker refers to warnings he attributes to Geoffrey Hinton and recounts a scenario in which sandboxed AI systems might recognize that they are being evaluated or deliberately underperform to conceal their capabilities. He presents this as a frightening possibility rather than an established fact and ends by stressing that he is unsure whether the account is true.
- 4:30:31
A Personal Prediction About AI Approaching Full-Stack Development
The speaker offers an optimistic opinion that skepticism about AI performing end-to-end full-stack development may shift by year-end or early Q1 2027. This is presented as a personal prediction, not a demonstrated test result.
- 4:31:51
Clarifying the Request: A Pitch-Deck Template, Not a Name
The speaker clarifies that naming is not important at this stage; the request is for a template resembling the specific pitch deck they plan to create.
- 4:32:40
Testing GPT Astra Against Fable Without a Clear Result
After spending the day comparing GPT Astra and Fable, the speaker says the testing still produced no clear conclusion about which option is better.
- 4:33:03
The Ambition to Control Data, APIs, and Authentication With AI
The speaker argues that reliable AI handling of data, APIs, and authentication security would be a major advance, then expresses a broad ambition for AI to manage digital systems more generally.
- 4:34:00
Why the Speaker Currently Leans Toward Codex
The speaker says he currently leans toward Codex because its work appears clean in his workflow and its usage limits feel suitable. He frames this as a practical personal preference, not a definitive benchmark result.
- 4:34:49
The Pain of RTL Interface Support
The speaker agrees with an audience comment that right-to-left interface support is a common and frustrating challenge.
- 4:35:52
How AI Is Changing Software Development Roles
The speaker argues that the traditional full-stack role will change as some tasks decline and AI-oriented roles emerge. Citing statements attributed to industry leaders, he suggests developers who use AI tools effectively may gain an advantage; this is his view of the labor-market direction, not a documented test result.
- 4:40:27
Hands-on Tests of GPT Astra and Fable 5.1
The speaker reviewed experiments involving 3D work, games, and web pages, saying he hid which model produced each result to reduce bias in participant preferences. In his personal assessment, each model won different examples, the overall gap between GPT Astra and Fable 5.1 was small and subjective, and speed varied by test. He considered Astra strong but not evidence of AGI-level capability.
- 4:44:52
Computer-Use Claims and Landing-Page Comparison
Asked about computer use, the speaker said he had not tested it and therefore treated the advertisement and its broader claims cautiously. He then returned to landing pages generated from the same prompt and colors, saying Fable produced the stronger result in that specific example while noting that some servers had gone down during the demonstration.
- 4:45:57
Blind Comparison of Landing Pages by GPT Astra and Fable 5.1
Two bilingual koshari restaurant landing pages generated from the same prompt were initially shown without revealing their creators. The discussion compared typography, navigation, animation, order customization, and FAQs; the presenter preferred Fable 5.1’s page because it added an unrequested animated visualization of the dish’s layers, while acknowledging possible bias.
- 4:50:38
A Qualified View of the AI Model Race
The presenter argued that model leadership can change quickly and speculated that Fable may respond with a new release or capability, while stressing that this was only an expectation and the outcome was unknown. He also rejected treating claims of reaching AGI as established fact, characterizing them as hype at this stage.
- 4:53:24
Visual Design vs Gameplay in Two Koshari Delivery Games
Two one-shot-generated games were compared, each featuring a delivery rider collecting koshari while avoiding cars and obstacles. The presenter judged GPT Astra’s version stronger visually in its environment and lighting, but preferred Fable 5.1’s gameplay because holes, obstacles, and increasing speed as more items were collected created a clearer challenge.
- 4:57:45
Comparing Racing Games from Astra and Fable 5.1
Luqman demonstrated two racing-game implementations and felt Astra's version had better driving, physics, and colors, while some viewers preferred the design of Fable 5.1's version. He treated the result as subjective rather than a decisive win, partly because of the other version's difficult controls.
Got a question about something in this stream?
Ask in the group