@iamcheyan: Recently we were doing data migration and needed to build a comparison tool to verify the migration results. I said I could do it, and Python would be very suitable for this. But after a round of discussions, they decided to implement it in COBOL instead, claiming they've been using it for over 50 years... The ancestral methods must not change.
Summary
When migrating data, the author proposed using Python to build a comparison tool, but the team insisted on using COBOL due to historical reasons (over 50 years of use), reflecting the inertia of legacy technology.
Similar Articles
@ai_super_niko: https://x.com/ai_super_niko/status/2077242414243480056
A experience sharing about migrating from Claude Code to Codex, introducing the process of automatic import configuration, the Command and local rules that need manual adjustment, and the division of labor between the two tools after migration.
@Fenng: Ali fully bans internal employees from using Claude Code in office environment. This news reminds me of the group of people who used Yiyuyan for Chinese programming ten or twenty years ago.
Alibaba has completely banned internal employees from using Claude Code in the office environment, sparking the author's association with the group of people who used Yiyuyan for Chinese programming back then.
@bozhou_ai: Can someone tell me, now that we already have codex, Claude code, what do we still need openclaw or Hermes for?
A discussion question comparing AI coding tools (Codex, Claude code, openclaw, Hermes).
@yibie: Databricks tested various coding tools with real tasks from their own team—conclusion: the same model called from different harnesses has a per-task cost difference of more than 2x, while quality is the same. "Pi: Minimalism and High Performance" Pi's minimalism is its advantage AI makes code cheaper, and as a result many companies build bigger...
Databricks' benchmark shows that the same model invoked through different harnesses has a cost difference of more than 2x, while Pi, as a minimalist coding harness, delivers high performance at low cost; Shopify also used Pi to extend Autoresearch and improve efficiency.
@mylifcc: Jason's comparison of real projects is well-written. Fable 5 indeed shines in complex coding tasks: almost no low-level mistakes, proactively considers edge cases, strictly adheres to the brand system, and can quickly solve tasks in a 13-year-old codebase where GPT has failed multiple times...
User @mylifcc shares their evaluation of Fable 5 and GPT-5.6 Sol on complex coding tasks, believing that Fable 5 has high accuracy but high cost, proposes a hybrid workflow model, sparking discussion on combining model usage.