AI Agents Move Into Billing Disputes, Video Editing, and Reverse Engineering as Builders Debate What Work Becomes

Key Takeaways

1. A Codex-driven agent spent about an hour fighting a Verizon billing failure and won a service hold, a concrete case of LLMs as bureaucratic dispute resolvers.
2. HyperFrames Studio drew a strong first-day review as an easy AI video editor, with real limits: one-line prompts produce junk, and the app still has ghost frames and clumsy timeline navigation.
3. Levelsio says his retro computing projects — Windows 3.1, Windows XP, Quake in WASM — went unnoticed for two years and now draw 80,000 daily browser players.
4. Noah Kagan hit 421,410 verified impressions on day 29 of a 500K challenge by repeating three fixed content formats instead of chasing new ideas.
5. Codie Sanchez's BizScout Score data shows over 50% of listed car washes cannot service an SBA loan at asking price, while the highest-scoring restaurants get ignored.
6. Arvid Kahl argues the AI debate skips UBI, and that people will still choose work in an AI future.
7. Acquire.com listings and commentary highlight how moats and multiples vary sharply by business model.

1. AI Agents in the Wild

1. Patrick McKenzie amplified a case in which a person stuck in a Verizon billing loop — told they owed money or would lose service, yet blocked from paying — turned to Codex desktop. The agent drove a browser against Verizon's broken web app for about an hour, used a password manager and email with authorization, and only worked after being switched to a high-end Astra model; in a long AI-versus-AI argument in Verizon's support chat, it got the other bot to admit a temporary backend problem from an in-progress billing platform update and to set a service suspension hold. The poster said they likely would not have thought to request that, especially as their emotional resources drained. McKenzie's comment: LLMs are good at playing "dangerous professional" against difficult, badly managed bureaucracies. — via 1
2. Ben Tossell called @bot his current favorite tool but said it keeps insisting it cannot access his MacBook, can occasionally be tricked past that, and usually cannot; he is still unsure about the form factor but finds the interface quite good. — via 1
3. Tony Dinh suggested ChatGPT should automatically compress context in the background as it nears the limit, instead of making users wait for compression each time. — via 1

2. AI Video Editing and Creator Tooling

1. Alex Lieberman reviewed HeyGen's HyperFrames Studio as the easiest AI video editing tool he has used, saying he produced a 25-minute podcast with intro, captions, and archival footage, plus table-to-video and multiple app launch videos, within 24 hours and without Premiere or Final Cut; the workflow is screen circling, text boxes, and natural-language commands. He described it as upgraded from a "geek tool" requiring prompts in Claude Code to an editor with a timeline, chat box, and brush, where you can circle a problem and dictate a fix, and said it is hard to go back to hand-written timestamps. The barrier is a Claude, Codex, or Grok subscription plus a Mac or Linux machine, and he offered a 15-minute walkthrough with six real-file build cases (CSV, voice memo, Luma link, 25-minute podcast, ambient footage, selfie video) plus pros and cons. He acknowledged limits: a single sentence yields "garbage content," the app is on day one with ghost frames and clumsy timeline navigation, though updates ship multiple times daily; target users include zero-experience newcomers, accountants and marketers relying on spreadsheets and decks, people burning tokens repeatedly on Higgsfield clips, and professional editors who want to hand off tedious tasks to AI. — via 1
2. Lieberman said that amid accelerating technology and rising pressure to learn AI, he tries to spend equal time studying how people work, storytelling and creative thinking, and trust-building; his weekend consumption included Jay Shetty interviewing Tony Robbins on mindset, David Perell interviewing Morgan Housel on nonfiction writing, a Dave Chappelle stand-up special, David Senra interviewing Ari Emanuel on building WME/TKO, and Gary Halbert's copywriting book The Boron Letters. — via 1

3. Retro Computing, WASM, and the AI Reverse-Engineering Wave

1. Levelsio recounted two years of nostalgia projects: he first wanted to run Windows 3.1 in a browser, which took months; then built a dial-up connection, extremely hard given the AI capabilities of the time, eventually done with @bai0. He also ported classics like Quake to WASM with multiplayer servers, with Quake III done by @lukathedev, and spent six months trying to reverse-run Windows XP on @copy_sh's v86, which was extremely hard due to size and only possible with modern AI like Opus 5.5. He said these projects got almost no attention for two years and are now rising with AI-assisted reverse engineering and WASM porting; his @pieter project is positioned as a non-commercial preservation project, like an interactive online computer museum, and yesterday 80,000 people played Windows 3.1 or Windows XP in a browser, chatted with MSN and AOL Messenger AI bots, and played 3D Pinball or Quake. — via 1
2. Levelsio also corrected his earlier misreading of X payout data: a notification said "You got paid," and he misread the $11K total figure; the Original Content and Revenue Sharing programs paid roughly comparable amounts, possibly slightly lower for one, depending on views. He said his video was among the highest-viewed content of the period, which earned @JohnONolan a lot of steak, and cited 64 million impressions for the period. He also reposted a claim that employee cost and productivity are negatively correlated. — via 1 2 3 4

4. Creator Growth, Business Data, and Work in an AI Future

1. Noah Kagan said day 29 of his "500K impressions" challenge reached 421,410 verified impressions, 78,600 short of the goal; the past week added 68,200, versus only 10,800 in the prior 10 days. He attributed the growth to abandoning creativity and repeating three fixed formats: attacking expensive software (a Granola post with 150K views), telling a founder friend's real numbers and lessons (a Sam Parr story with 65K views), and giving copyable prompts plus results (a Claude prompt with 60K views and 1,409 saves). His failures were a morning scorecard (5,000 views) and an Amazon tip (110K views), which lacked copyable content or drifted off-topic. His advice: find your three formats and repeat them until bored, acknowledging it is less fun and more like work, but "boring wins." — via 1
2. Codie Sanchez said her marketplace launched a BizScout Score this year that compares each listing with peers on a 0-100 scale, took nearly four months to build, and is now the primary way buyers sort listings on BizScout. Disclosed insights: over 50% of listed car washes cannot service an SBA loan at asking price; buyers view laundromats 3.5 times as often as hair salons, but salons are much cheaper relative to cash flow; manufacturing businesses see 4x more inquiries when scored high, while the highest-scoring restaurants are completely ignored. On AI, she said she initially thought ChatGPT would destroy many jobs, but after observing usage now thinks its output is poor and worse than not using it; her contrarian view is that AI is not a threat provided you truly care about the work and have taste and curiosity, traits that will be worth more than ever. — via 1 2
3. Arvid Kahl criticized elite gatekeeping in mathematics — the idea that math should only advance as professional mathematicians see fit — arguing mathematicians do not own mathematics. He said the current AI debate ignores universal basic income: mathematicians fear progress because funding supports their families, while developers welcome AI writing code but fear losing their own jobs, so UBI should be demanded. He said using AI counts as both "working hard" and "barely working," and invoked Star Trek to argue that in a future where AI replaces existing jobs, people will still choose to work because life is more than commuting and agile retrospectives. — via 1 2 3 4
4. Acquire.com listed a scalable creative design agency on monthly subscriptions with $239.8K TTM revenue for $253.8K; a white-label student transportation SaaS with GPS and compliance features at $181K TTM profit and $218.7K TTM revenue for $363K; a roughly 0.4-year-old TikTok partnership agency at about $27K MRR and $196K TTM revenue for $199K; and a turnkey e-commerce growth agency, SOP-driven and high-margin, at $55.4K TTM revenue for $50K. It also shared Ethan Roberts' story: while working full-time in cybersecurity, he found an unsolved problem in wedding planning software, built PlanEase, got first customers via Facebook groups and cold email, listed on Acquire.com, found a buyer, and is now self-funding another startup with part of the proceeds. The account's view: moats vary greatly by business model, and buyers compare "how hard to replace" rather than just valuation multiples. — via 1 2 3 4 5 6
5. Starter Story relayed a no-coding-experience founder's method for building a $50K/month app in six months: hire low-cost virtual assistants on Fiverr, screen and cold-DM many creators against four criteria — 5,000 to 50,000 followers, consistent posting, on-camera presence, and active comments. It also offered a "change your life before 2027" routine: watch one Starter Story episode, pick an idea and reverse-engineer it, set a two-hour timer and start, then repeat step three daily until 2027. — via 1 2
6. Justin Welsh amplified the view that you do not need a $10M business, but should build multiple small income streams that steadily convert attention into $1,370 a day; if you avoid lifestyle creep, $500K a year lets you do almost anything. Another amplified view: ultimate success by 45 is a solid marriage, a healthy body, financial freedom, and needing to please no one. Welsh said people mock those who change positions over time, but what truly scares him is people who never evolve. — via 1 2 3
7. Codie Sanchez argued interpersonal contagion is underrated and urged ruthlessly removing negative people, citing Federer changing dentists over excessive negativity. Sam Parr praised Matt's transparency as worth learning from, said overall US wealth is extremely high and ordinary Americans live well with air-conditioned houses, big cars, and cash, and that millionaires are numerous, while conceding housing is still too expensive and other areas need improvement; he cited data that 80,000 households have over $100M in assets, guessed only about 10% truly have $100M investable, strongly rejected the claim that ordinary Americans cannot cover a $500 emergency as completely false, and noted he is not sure the cited data is accurate. Nick Huber said parenting and managing people have much in common and shared how to raise children into capable decision-makers. — via 1 2 3 4 5