Reading up on IBM

A vacant IBM campus in Somers, N.Y. has become a destination for trespassing urban explorers, drawing police responses and local frustration.

A Linux Foundation working group proposes DocLang, an XML-based document format optimized for LLM tokenizers to reduce cost and improve accuracy in enterprise AI document processing.

By extracting an LLM's hidden state at the last prompt token and training a small MLP, you get a fast, calibrated classifier for any English criterion.

A 1964 IBM training video explains the benefits of direct-access storage devices (DASD) over index sequential access (ISAM), using a flipchart presentation.

Microsoft released the earliest known DOS source code (86-DOS 1.00 and PC-DOS 1.00 snapshots), predating MS-DOS branding, to preserve computing history.

The Trump administration is awarding $2 billion in grants to nine quantum-computing companies, including IBM and GlobalFoundries, in exchange for U.S. government equity stakes.

A Microsoft engineer in Boca Raton escalated a TAB-key dispute to his manager, who refused to overrule him, leading to an IBM VP opposing the choice and a deadpan reply ending the debate.

Hacker News commenters argue career growth is self-driven, not a manager's responsibility, while criticizing Amazon's lack of mentorship.

Anthropic's blog post suggesting Claude Code can rewrite COBOL apps contributed to a 13% IBM stock drop, despite IBM itself promoting similar AI tools since 2023.

COBOL developers find AI coding tools helpful for documentation and simple tasks but not yet a threat due to legacy system complexity and compliance constraints.

The massive overinvestment in LLM datacenters is unsustainable and will lead to a bust that makes GPU supercomputers cheap, fueling a new science and technology revolution.

A blog post argues that many current tech trends (AI surveillance, locked-down phones, enshittification) were not inevitable but resulted from specific choices that can still be rejected or reversed.

Hacker News commenters react to IBM's acquisition of Confluent with widespread skepticism, citing IBM's history of poor acquisitions and suggesting Kafka alternatives.