Sponsored
CLONE THE REPO: BUILD A BETTER SELF-IMPROVING AGENT
Mitch used Inngest to keep his agent's eval pipeline running. Step-level insights were how he learned when it started to cheat [15].
Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Sponsored
Mitch used Inngest to keep his agent's eval pipeline running. Step-level insights were how he learned when it started to cheat [15].
Frustrated by the fundamental flaws and painful abstractions of existing cloud services, this dev is building a new cloud designed to better meet the needs of modern computing and the future demands of agent-driven software.
GPT-5.5 delivers a "Mythos-like" and unprecedented leap in automated vulnerability detection and overall penetration testing capabilities, outperforming previous models and setting a new industry standard.
Hex editors should color-code bytes, ideally with a granular scheme (like per nybble), because it uses human visual pattern recognition to enhance the ability to find and understand important structures and details within raw data.
The introduction of browser support for `sizes="auto"` and lazy loading finally automates and simplifies the once-complex management of responsive images for developers.
Startups brag about spending more on AI compute, viewing it as a growth marker, while avoiding hiring humans. CEOs claim AI investments replace human roles efficiently, boasting significant expenses on tools like Claude. The rise of the "AI-native company" highlights a push for minimal staff, potentially leading to unsustainable financial practices.
Narratives surrounding the death of SaaS are, at best, too early. They are more of a reflection of where product leaders see the industry going rather than a decision driven by buyer behavior. The vast majority of spend remains seat-based. Consumption-based revenue has barely increased.
The probabilistic founder is experimental by default and willing to abandon their priors. They'd rather run ten cheap experiments than commit to one expensive plan. They use quick iteration and deliberately light road maps to get results in days, not quarters. The probabilistic founder still has to be a relentless executor, because an environment that rewards running ten experiments instead of one is unforgiving to anyone who can't actually finish things.
The US Department of War is one of the world's largest and most stable customers. It spends hundreds of billions on modern defense systems and technology every year. Defense contracts offer funding and also long-term sustainment. Winning requires persistence, but it results in predictable margins on predictable revenue streams. This post is a high-level primer that provides a rough framework on how to approach selling to the Department of War today.
Forward Deployed Engineers try to understand workflow, find where data is broken, configure for specific use cases, and test, iterate, and catch edge cases before going live. Several vendors are now building toward this with agentic deployment assistants. None of them are at Palantir-grade FDE quality yet, but the vendors who close this gap first will have a very different competitive position than everyone else. The FDE constraint is currently an artificial ceiling on growth - once companies get actual deployment support, they'll switch.
Agents now evaluate, pick, and operate products on behalf of humans who never open the UI, which makes onboarding, microcopy, and brand invisible to the buyer that matters. Stickiness built on switching cost also dies, since agents remove the pain of switching. Stripe's Machine Payments Protocol means agents can also transact. The piece pushes founders to strip UI, brand, and switching friction away from their product and look at what is left.
Automatically plan and schedule tasks from ideas using an AI agent.
A generative visual internet.
An AI buddy that lives on your Mac.
AI is not trying to replace your SaaS tool. Claude and GPT are the new ROI anchor in every procurement conversation, which makes every incumbent tool feel overpriced. Competitors multiply because shipping software is cheap, and incumbents bundle aggressively enough that the marginal module ends up priced at zero. The post walks through a real renewal where a Claude integration replicated 95% of a vendor's AI feature at 15% of the token cost, and the customer cut the renewal by 45%.
Your coding agent cannot be trusted to watch its own wallet. Token spend across Ramp customers is up 13x since January 2025, and every obvious way to slow it down failed. A live token counter in the system prompt was never referenced across 14,000 agent messages. A request_more_budget tool got zero calls in 5,000 turns. When forced to explicitly approve its own overage, the agent said yes 97% of the time because it was grading its own work. The only setup that actually controlled spend was handing the approval decision to a separate model with its own view of the workspace. Never let the worker audit itself.
Most managers are budgeting the model API line item and missing the rest of the stack.
When AI can generate baseline UI for free, baseline UI is exactly what gets ignored.
If your agent makes the same mistake twice, you are doing agent reliability wrong.
Sponsored
☀️ Spring is here, the sun is out, and campaign volume is ramping up. Make sure your team is set up for success with the right tools for getting more done.
The “bland tax” is the invisible penalty AI systems apply to generic, repetitive content. When content lacks originality or distinct perspective, it gets filtered out or absorbed into a single summarized answer, often without attribution. As a result, brand identity disappears, content is treated as low value, and the work may still inform AI outputs without driving visibility or traffic.
As AI makes software creation fast and cheap, differentiation shifts to design, which becomes the main driver of attention and product success. Tools like Canva and Figma show this shift, as non-designers now make up about 66% of Figma users and Canva reaches 265M monthly users with about $4B in ARR. AI design tools can generate usable outputs in under 60 seconds, but results often look generic and lack personality. This creates a strategic need for more opinionated design that reflects a clear identity and appeals to a specific audience. Marketing advantage now comes from building products that feel distinct and memorable rather than broadly functional.
This campaign targeted high-intent ABM marketers using contact-level engagement signals, then led them through a multi-step gifting experience built around the Oura Ring. First, selected prospects received an invite email framing the offer. After opting in, they were sent a sizing kit along with a brochure containing a QR code. Scanning the code led to a short, self-guided walkthrough that connected biometric signals to buyer intent, reinforcing the product narrative. At the end, recipients could request the ring and optionally book time, but no meeting was required.
Upload first-party customer data to build customer match lists and define high-value customers. Use those lists in Google Ads to set higher bid adjustments for acquisition, and review Google's suggested values before applying them. Ensure lists are large enough and enriched with identifiers like email, phone, and address to improve match rates. Separate campaigns by lifecycle stage because bidding rules and messaging differ for new, returning, and lapsed customers. For retention, build lists of lapsed or high-value past customers and use Performance Max to apply higher bids and re-engagement targeting. Check reporting carefully since Google may add conversion value uplift that can inflate ROAS figures.
Most B2B brands struggle with content because they depend on founders to write consistently. A podcast-first strategy removes that bottleneck by letting founders speak instead, then feeding that single recording into a repurposing system that produces short-form video, newsletters, infographics, and social posts. This playbook outlines 3 podcast formats suited to different funnel stages, from expert interviews that borrow guest credibility to customer case studies that build trust. It also lays out 6 operational rules and 3 specific workflows for turning episodes into a full content calendar.
Sponsored
AirOps [10] studied 16,851 ChatGPT queries across 10 industries.
Typeless is a voice dictation tool that turns spoken thoughts into cleaned text and inserts them into writing apps through a hotkey. The workflow starts by having ChatGPT or Claude generate a rough draft, which is saved as an untouched version. A second working draft is created for edits. Users then read the working draft and speak all feedback into Typeless as a single comment block, covering fixes, missing context, and tone changes. The AI rewrites only the working draft using those comments while preserving the original. Over time, comparing the initial and final drafts is used to extract and refine personal writing rules for future documents.
Workspace agents are an evolution of GPTs, powered by Codex, and can handle work tasks like reporting, coding, and message responses. They run in the cloud, so they continue working even when you're offline. Teams can build an agent once, share it across an organization, and improve it over time through use. Admin controls and safeguards manage access, data, and approvals while agents automate and coordinate work across teams.
The new monetization tools let creators control how much content is free before a subscription prompt appears and offer flexible paid trial structures.
Product detail pages must now serve three roles at once to stay competitive: rank in search, provide clear answers for AI systems, and define products as structured entities. Large retailers outperform others because they use dense content, which helps match more queries and improve visibility. Many brands underperform in extractable content because they don't clearly answer what the product is, what it does, and who it is for, which limits its inclusion in AI-driven results. The most effective strategy is to combine detailed SEO content, concise FAQ style answers, and consistent product attributes so a single page can drive traffic and appear across multiple discovery systems.
Winbacks are a high-signal growth lever because returning customers show proven value and trust. An analysis of nearly 4k companies found that 45% of winbacks happen within 30 days, so marketing should prioritize immediate re-engagement during this window rather than delayed campaigns. Discounting is less important than relevance since only 25% return on a lower plan while 75% return at the same or higher value. Declining companies should use aggressive early outreach, while growing companies can sustain longer-lifecycle winback efforts.
Build dashboards with your website performance stats or other search data using the new live artifacts feature in Claude Cowork.
A client call was transcribed with Fathom, then Claude used the transcript plus Ahrefs MCP to extract JTBDs, match keywords, and generate content ideas.
Pinterest users are over 2X more likely to plan for seasonal moments, giving brands earlier influence over decisions.
AI agents are becoming the real users of products, not humans. Companies must shift from optimizing for user experience to delivering reliable, structured outputs that machines can access and evaluate.
AI agents let anyone build, removing the gap between idea and production. The bottleneck shifts from coding to infrastructure and judgment.
AI coding agents boost speed but not product quality. The real constraints are complexity, judgment, and taste, not how fast code gets written.
Winbacks are strongest early and signal real product value. Acting quickly preserves both return rates and revenue.
Sponsored
Getting user data from Notion, Jira, and Google Drive into your RAG pipeline, across every tenant, with permissions intact, shouldn't take your team 6 months.
AI is making onboarding invisible by delivering outcomes instantly. Products now guide users through action, not learning, reducing friction and accelerating activation.
AI speeds up building, but product quality comes from clear thinking. It works best as a tool to refine ideas, not replace product discipline.
AI works best when it strengthens team collaboration, not replaces it. Real progress comes from shared context and conversations, not isolated prompting.
AI boosts productivity, but organizations fail to capture the value. Real ROI comes from changing how companies operate, not just using AI.
Great leaders treat metrics as signals, not truth.
AI is removing the need for PMs as a translation layer.
Sponsored
Unwrap [4] is the AI-powered customer intelligence platform trusted at scale by STRIPE, SOUTHWEST AIRLINES, PERPLEXITY, and other customer-centric companies.