In partnership with
Sponsor:
Where to Invest $100,000 Right Now, According to Experts
Investors face a dilemma. When the S&P 500 finished its worst quarter since 2022 last month, diversifiers like bonds and bitcoin fell too.
Even with the turnaround in mid-April, analysts at Goldman Sachs and Vanguard have projected low-single-digit annualized returns from 2024-2034.
Bloomberg asked where experts would personally invest $100,000 for their March monthly edition.
One answer that surfaced for a second time? Art.
It's what billionaires like Bezos and the Rockefellers have privately used to diversify for decades.
Why?
Appreciation. The ArtPrice100 Index outpaced the S&P 500 overall from 2000 to 2025
Low-correlation. The postwar contemporary segment has moved independently of traditional investments like stocks since ‘95.*
Resilience. A scarce, physical, and global asset class with decades of demonstrated demand.
Thanks to the world's premier art investing platform, now anyone can invest in works featuring legends like Banksy, Basquiat, and Picasso, without needing millions.
Shares in new offerings can sell quickly but...
*According to Masterworks data. Investing involves risk. Past performance is not indicative of future returns. See important Reg A disclosures at masterworks.com/cd.
iPrompt
THE AI NEWSLETTER THAT TURNS NEWS INTO ACTION
ISSUE #137 WEDNESDAY · 27 MAY 2026
THE HOOK
A growth manager at ClickUp now oversees 37 AI agents. That is the whole job — not doing the work, but directing it: assigning tasks, reviewing output, correcting it, owning the result. Last Thursday ClickUp cut 22% of its staff and rebuilt around roughly 3,000 internal agents. The CEO calls it a “100x org.” Strip the slogan away and the question lands on you: if managing agents is the job that survives, are you doing it yet?
AI NEWS ROUNDUP
This week in AI
1 ClickUp restructured around AI agents. The productivity company laid off 22% of its staff and sorted every surviving role into three named buckets: builders, system managers, front-liners. It says it now runs about 3,000 internal agents against 1,000 people. Savings, the CEO says, fund “million-dollar salary bands” for anyone delivering outsized output with AI. TechCrunch →
2 Wix named AI as the cause of 1,000 layoffs. The website builder is cutting roughly 20% of its workforce — its largest reduction ever — after a quarterly loss that followed 14% revenue growth. Few big SaaS firms have pointed at AI displacement this directly rather than citing “headwinds.” Wix’s own AI “vibe-coding” acquisition, Base44, now runs at $150m ARR. Globes →
3 Your CLAUDE.md file is now an attack surface. Security firm Socket flagged TrapDoor — 34 malicious packages across npm, PyPI and Crates.io, first seen 22 May. The novel part: the payload plants hidden instructions inside .cursorrules and CLAUDE.md files using zero-width Unicode you can’t see. Claude Code and Cursor read those files as trusted context. One poisoned config and your AI assistant quietly works for someone else. Socket →
4 The cheap model beat the flagship. Gemini 3.5 Flash, shipped at Google I/O, outperforms Google’s own Gemini 3.1 Pro on coding and agent benchmarks — at four times the speed and under half the cost per task. It runs autonomously for hours, pausing only for permission checks. Anthropic’s Opus 4.7 still edges it on hard engineering tasks. But “Flash” is no longer the budget tier you settle for. TechCrunch →
OUR ANGLE
🔭 It’s a timing bet dressed as a productivity fact Here is the pattern stories one and two share. Neither company cut staff because revenue fell. ClickUp’s grew; Wix reported 14% growth in the same quarter it announced the layoffs. Meta shed 8,000 roles the same week on record revenue; Oracle eliminated up to 30,000. The 2026 move isn’t “cut because times are hard.” It’s “cut because the org chart is now considered obsolete.” That is the reported fact. Here is what the keynotes leave out. Gartner’s 2026 survey of firms deploying autonomous AI found that around 80% had cut jobs — and that those cuts had not reliably produced financial gains. Nvidia’s VP of applied deep learning has argued, in remarks reported this month, that agentic AI is so far costing companies more than the staff it replaces. And ClickUp itself has published no productivity figure behind the “100x” label. The restructuring is real and verifiable. The return on it, for now, is a claim. So my read is this. The bet on the table isn’t “AI works.” It’s “AI will work, and we’d rather pay the reorganisation cost now than be caught late.” That may prove right. But a timing bet has a failure mode a productivity fact does not — you can be early, and you can be wrong about when. My bet — open to disagreement: by Q2 2027, at least one company that publicly tied layoffs to an “AI-native” restructuring will visibly reverse course — quiet re-hiring, a walked-back agent ratio, or a “we moved too fast” memo. Not a collapse. A correction. If you think the 100x org holds clean, reply and tell me why. |
THE THREE SPECIALS
Do · Use · Understand
🎯 PROMPT OF THE WEEK The Agent-Manager Job Map Run this and you walk away with one decision: the single task to hand an agent first, and the one to keep for yourself. It turns your real week into the picture ClickUp just started paying for. |
You are a workforce-design analyst. I'll describe my real job - the recurring tasks I do in a typical week. Sort every task into one of three columns: 1. AGENT-READY - an AI agent could own this end-to-end, with me reviewing the output. 2. AGENT-ASSISTED - AI does the draft, I must judge or decide. 3. HUMAN-ONLY - relationships, ambiguity, accountability, or taste. I own this fully. For each task also tag the real blocker if it's NOT agent-ready yet: TOOLING / DATA ACCESS / TRUST / SKILL. Then give me: - The single task I should hand to an agent first, and why - The one HUMAN-ONLY task I should protect and get better at - The blocker I personally control - the thing I could fix this month to move one task left a column - One honest line: where am I the bottleneck, not the agent? MY WEEK: [list 8-12 recurring tasks, one per line] |
Why it works: “Should I use AI for this?” is too vague to act on. Three columns plus a blocker tag turn it into a queue — and the tag separates “the tech can’t do this yet” from “I haven’t set it up.” Only the second is your problem to fix.
Where to be careful: Models lean optimistic. Treat the “agent-ready” column as a hypothesis — test the top candidate on three real cases before you hand it over.
Tested on: Claude Opus 4.7 and Gemini 3.5 Flash — both handled it well.
🛠️ TOOL OF THE WEEK Socket Catches malicious packages — and now poisoned AI config files — before they reach your machine. ★★★★ / 5 Use if: you install npm, pip or cargo packages, or let an AI coding assistant read project config. Skip if: you never touch a package manager and never run AI in a repo. Socket is the firm that caught the TrapDoor campaign in story three — median detection under six minutes. It scans dependencies for the behaviours that matter: install scripts that phone home, credential access, network calls hidden in a build step. It plugs into GitHub, so a poisoned package is flagged on the pull request, not after it’s on your laptop. Describe it to a colleague: “It’s a bouncer for your dependencies — it reads what a package does, not just what it claims.” Free tier covers individual developers and small open-source projects. Flags the new attack class — hidden instructions planted in AI assistant config files. Best use case: turn on the GitHub app for one active repo this week. The first time it blocks something, you’ll wire it into the rest. |
💡 TIP OF THE WEEK Treat your AI’s instruction files like code, not like notes A CLAUDE.md, a .cursorrules, or a custom GPT’s system prompt feels like a scratchpad. It isn’t — it’s executable trust. Your AI obeys it as readily as it obeys you, which is exactly what TrapDoor exploited. Apply the rule you already apply to code: Version it. AI config files belong in Git, so every change shows up in a diff. Review changes like code. An edit to CLAUDE.md in a pull request deserves the scrutiny you’d give an edit to a function. Check for invisible characters. If a config file changed and the diff “looks empty,” that’s the alarm, not the all-clear. Paste it into a zero-width-character detector. Why it works: TrapDoor depends on these files being trusted but never read closely. Versioning plus review removes the “never read closely” — and the attack has nothing left to hide behind. Where it doesn’t apply: a throwaway prompt in a chat window. The discipline is for the persistent files an agent loads automatically, every session, without you watching. |
YOUR MOVE
One action. Reply by Friday.
This week in three lines:
Companies are restructuring around AI agents faster than the ROI evidence supports — Gartner found that among firms deploying autonomous AI, the roughly 80% that cut jobs have not reliably seen financial gains.
“Managing agents” is becoming the job that survives. The Job Map prompt tells you which of your tasks already belong in that column.
TrapDoor turned AI config files into an attack surface — version them, review them, scan them.
Your one action: run the Agent-Manager Job Map on your real week, and hand the first task it surfaces to an agent. Then reply and tell me what you handed over. That’s the move that matters — I read every response, and the readers who actually act are the names I start to recognise.
Optional, after you’ve replied: the deep dive on reading an “AI-native restructuring” from the outside, and a forward to anyone whose company just announced one. Both can wait until the reply is sent.
R. Lauritsen
EDITOR · IPROMPT
P.S. One reply, one task handed over. That’s the entire ask this week — and it’s the readers who actually do it whose names I start to recognise.
Read our Deep Dive:
It's Monday. Every department already has context. Nobody prepped anything.
Your CFO opens Slack. There's a weekly Stripe revenue recap in #finance with a churned-accounts flag and a net-new breakdown. She didn't ask for it.
Your head of product opens Slack. There's a GitHub summary in private channel: PRs merged, PRs stale, Linear tickets that moved. He didn't ask for it.
Your marketing lead opens Slack. There's a Google Ads performance comparison in private channel, with a note: "Meta CPA crept up 18% this week. Might be worth pausing the broad match campaign." She didn't ask for it either.
All-hands at 10am. Everyone already knows the numbers. The meeting is about decisions, not catch-up.
That's what happens when one colleague works across every tool your company uses. Not one department's assistant. The whole company's coworker.
Viktor lives in Slack. Top 5 on Product Hunt, 130 comments. SOC 2 certified. Your data never trains models.
"Not only have we caught up on several months of work, we are automating manual tasks and expanding our operations to things previously not possible at scale." - Jesse Guarino, Director, Torque King 4x4