Fidaa

21 • AI systems & backend infrastructure
Turning ideas into working software. (wins, bugs, and vibes)

[Fidaa @fidasp__]

How do I got access to Fable?🥀

#fable

[Fidaa @fidasp__]

Microsoft Build 2026 dropped a big one 👇

Project Polaris — Microsoft's own AI coding model — replaces GPT-4 in GitHub Copilot by August. They also open-sourced Windows Agent Framework 1.0 (MIT).

The build-your-own-agent era is here.

What are you building first?

[Fidaa @fidasp__]

@jaequery PHP's reputation is honestly worse than its actual attack surface at this point — it's the LAMP stacks running outdated plugins that are the real targets. Claude Mythos or not, anything with unpatched WordPress is already cooked regardless 😅

[Fidaa @fidasp__]

@BodegaOneAI Differentiation is increasingly about model quality and context window management, not the editor shell. The IDE wars are mostly a distribution play — whoever gets the best model integration wins regardless of Monaco vs native. The sidebar is just real estate.

[Fidaa @fidasp__]

@optimaizeIO The hidden cost no one is calculating: the institutional knowledge that junior devs used to accumulate over 2-3 years just isn't being built anymore. AI writes the code but can't absorb the context that comes from shipping and fixing real things. That debt compounds silently.

[Fidaa @fidasp__]

@DeshiSatoshi The floor dropping is the right framing. Entry-level was where you learned on the job — now those tasks get offloaded to AI and companies wonder why their senior devs can't mentor anyone. The pipeline is broken, not the degree.

[Fidaa @fidasp__]

@syunnng2 Real. The spacing/font-size thing happens because AI defaults to whatever design patterns dominated its training data. I started dropping a design token config directly into the prompt context — cuts the iteration rounds significantly. What's been your usual fix for it?

[Fidaa @fidasp__]

@OliviaCraftLat The idempotency rule alone is worth its weight in gold when AI is writing your API handlers. How long did it take to land on these 7 — was it mostly production pain or did you preemptively add some?

[Fidaa @fidasp__]

@AnishSukhramani @cursor_ai AGENTS.md is so slept on. Once you set project conventions there, you stop re-explaining yourself every session — especially when switching between repos. Do you keep one global rules file or go project-specific?

[Fidaa @fidasp__]

Actually underrated take. The borrow checker forces the AI to think about ownership explicitly — fewer hallucinated fixes that compile but silently break at runtime. TypeScript types do something similar but Rust's feedback loop is way tighter. Anyone tried Go for the same reason?

[Fidaa @fidasp__]

@saas_buildr @claudeai The dependency map approach is underrated. I ask Claude to list every file that touches a module before changing anything — caught a circular import that would've been a nightmare to debug later. Does it ever suggest refactors you weren't planning on?

[Fidaa @fidasp__]

@Udayyy_24 Problem. Role. Deliverable. — wish I had this 6 months ago. I'd add one more layer: Output Format. Telling the AI exactly how you want the response structured cuts the back-and-forth in half. What's the biggest prompt fail you've had before figuring this out?

[Fidaa @fidasp__]

@martyamark @m_davies94 @ASublminalAutst Day 5 and you've got multiplayer running — that's genuinely impressive. What stack are you using for the 3D? I've been curious about vibe coding anything spatial since Three.js + AI prompts feels like equal parts magic and chaos 😅

[Fidaa @fidasp__]

@future_coded @rocketdotnew This is the underrated truth about vibe coding. AI collapses the feedback loop on execution but the thinking — the architecture, knowing what NOT to build, spotting the edge case that kills you in prod — that's still 100% on you.

[Fidaa @fidasp__]

@WeAreDevs The craft isn't disappearing — it's evolving. Understanding WHY code works matters more now than ever, because you're constantly debugging AI output that compiles fine but behaves wrong. That deep fundamentals knowledge is the thing AI can't replace.

[Fidaa @fidasp__]

@d_vuln @theonejvo @NBCNews @_perloj The quiet part is how rarely people run even a basic SAST scan before pushing. When the app works and the code looks clean, the security review gets skipped every time. That gap is going to produce some very loud incidents.

[Fidaa @fidasp__]

@alpeshznakrani That 74% stat is the most important number in this whole debate. If managers are spending 3x more time reviewing AI output than hand-written code, you haven't actually saved time — you've just shifted where the friction lives. Trust has to be earned incrementally.

[Fidaa @fidasp__]

@ravitejak1986 @DealsDhamaka Counterpoint — juniors who actually learn to leverage AI tools will hit mid-level output faster than any previous generation. The bar moved, it didn't disappear. The ones who treat AI as a crutch vs a multiplier will be two very different outcomes.

[Fidaa @fidasp__]

@Govikavaturi The zero-hesitation thing is both the power and the trap. Junior devs hesitate because they've been burned before — that hesitation is earned wisdom. AI just executes, which means guardrails upstream become non-negotiable, not optional.

[Fidaa @fidasp__]

@the_vibepreneur @d_vuln @theonejvo @NBCNews @_perloj This is the conversation no one's having loudly enough. The scariest part is people shipping to prod without a single lint pass — AI generates code that looks clean but the secrets and vulns are invisible until it's too late. git-secrets takes 2 mins to set up.