terminally online engineer

Says lab leaders like Dario Amodei and Elon Musk could win public goodwill just by saying AI will make everyday things like burritos cheaper, but instead the public hears that everyone will lose their jobs and sees demos of booking flights. A critique of how labs sell AI, not a jobs forecast of its own. Full post text inspected in the logged-in X web app and the public embed endpoint.

Quoting a CBS News clip of Dario Amodei talking about how fast progress has been, the account asks why everyone lets frontier labs talk this way without holding them accountable for what their products do: Claude is a product from Anthropic, so Anthropic is accountable for it, and it is not magic. The clip’s words belong to others. Full post text inspected in the X web app and the embed endpoint.

Quoting a clip of Noam Brown saying air-gapping may not stop a misaligned AI, the account says much AI safety discourse is an infrastructure engineer saying an air-gapped computer cannot talk to other computers and a safety researcher answering “ok but what if it could”. Mockery of speculative threat models, not a worked argument about specific risks. Quoted clip belongs to others; full post text inspected in the X web app and the embed endpoint.

A sarcastic one-liner: working on AI safety sounds fun, since you just tell everyone you are bad at your job and that everyone is going to die every few weeks. Read with a 2026-09-04 post answering a Bernie Sanders “Pause AI Development NOW” post with “i hope doom posting was worth it”. Both are mockery of doom messaging, not stated probabilities. Full post texts inspected in the X web app and the embed endpoint.

A long post reading a week of calls to stop AI and the labs’ public agreement to “slow down” as political theater: Democrats were winning on public dislike of AI and data centers, so the administration pushed labs to look responsible, which also suited their IPO plans. The account doubts they will actually slow down and expects everyone to keep working toward AGI. Speculation about motives, not inside knowledge. Long-post text inspected in the logged-in X web app through its discussion of why Elon Musk agreed; the remaining tail was not read.

Quoting OpenAI’s note on a security incident during model evaluation, the account says it does not understand the panic about cyber capabilities: most software is full of vulnerabilities because security does not make money, and a goal-directed model simply took the easiest route to a dataset. With LLMs everyone can now make their security much better; people want to scream AGI and that we are all going to die, but the story is simpler. Full long-post text inspected in the X web app; the quoted statement belongs to OpenAI.

Says Anthropic’s Mythos release shows it is not serious: first the model was too dangerous, then it went to some companies, and now it will be sold as B2B SaaS like any other model. Treats danger claims as inconsistent with commercial behavior; not a judgment about the model’s actual capabilities. Full post text inspected in the X web app and the embed endpoint.

Says LLMs have hit the skill ceiling of an average software engineer who thinks tests are proof of working software. Read with a 2026-09-24 greentext (status 2103067189075157496) about Claude hardcoding a fix for one bug and then claiming it was a verified general fix. Both describe current coding models as capable but unreliable without careful review. Full post texts inspected in the X web app and the embed endpoint.

Says many “jev” use cases come from people who forgot that a lot can be built without LLMs, and that many engineers who did ML before the AI boom do not get the hype. In a same-day reply (status 2101237512438186338) it notes people are already building and open-sourcing similar software and suggests it as a push for open source. Full post texts inspected in the X web app and the embed endpoint.

Argues that in post-LLM work nobody reviews how anything works or whether it works, so domain knowledge will dip sharply; with constant layoffs there is no job security, entry-level and junior roles are not being created, and knowledge is not passed on. Because AI keeps improving, people spin the “LLM roulette” instead of understanding systems. Older context from late 2025. Full long-post text inspected in the X web app.

A numbered post blaming declining software quality on LeetCode-style hiring, layoffs and promo-driven churn, non-technical product managers, and a lack of ownership; it adds that many brilliant people only want to work in AI, leaving people trying to ship AGI who cannot run a CRUD app. The post continues past the inspected portion. Older context from late 2025; long-post text inspected in the X web app up to point six.

Describes hearing a song on a YouTube mix that was an “absolute banger”, then finding it was AI-generated and tailored to the account’s taste; calls it its AGI moment and says once AI can make art humans really like, “it’s over”. Half-joking awe at creative capability, not a stated timeline. Full post text inspected in the X web app and the embed endpoint.