Insights

Substack AI Studio

2025-11-28

Stop paying for a studio: 10 prompts for a great LinkedIn photo

Practical tutorial to create a professional LinkedIn profile photo with Google AI Studio (Build mode) using image editing and the 'Nano Banana' model to preserve facial features. It includes a step-by-step workflow (upload a selfie, paste a prompt, iterate) plus 10 ready-to-run prompts for executive, tech, and creative styles focused on realistic lighting and framing.

• Edit a selfie (not generate from scratch) to keep identity consistent.
• In the prompt, specify lens/light, background, wardrobe, crop, and constraints (no filters, no caricature).
• Start from the 10 templates and iterate in small tweaks until it looks natural.

Read on Substack

LinkedIn AI Agents

2026-02-04

Claude Skills without code: importing skills via Manus AI

How to use Claude Skills without touching code: import a skill from GitHub into Manus AI and run slash commands like /seo-audit. It points to public skill repositories, highlights the crucial tip (copy the folder URL that contains SKILL.md), and walks through the import flow in the Manus UI.

• Copy the URL of the specific skill folder, not the repo root.
• Install in minutes via the UI and reuse it as commands in chat.
• Example: run an SEO audit tailored to a target audience.

View on LinkedIn

LinkedIn Privacy

2026-01-19

Confer: a privacy-focused alternative to ChatGPT

Introduces Confer (confer.to), a privacy-focused ChatGPT alternative launched by Signal's co-founder. The post claims signup generates a key/passkey you store in your password vault, and conversations are encrypted from there. It notes the higher entry price (US$35) and that it uses open-source LLMs without naming them.

• Privacy-first positioning: local key + encrypted sessions (per the product claim).
• Trade-offs: higher price and limited transparency about models.
• Worth evaluating for sensitive workflows (legal, compliance, internal data).

View on LinkedIn

LinkedIn Leadership

2026-01-11

Linus, AI, and leverage: efficiency beats purism

A real Linus Torvalds commit quote becomes an argument for AI as leverage: on specific tasks (e.g., visualization), the machine can outperform doing it by hand. The leadership takeaway is pragmatic: it is not about replacing expertise, but about increasing throughput and quality where it makes sense.

• Treat AI as leverage on narrow tasks with clear quality bars.
• Build a culture where 'better than by hand' is acceptable.
• Define criteria for when AI helps vs when risk/complexity is too high.

View on LinkedIn

LinkedIn Tech Trends

2026-01-01

Tech in 2026: what to expect (NotebookLM + Deep Research)

A 2026 tech outlook compiled in NotebookLM from top consulting firms, major tech companies, and deep-research sources. Key themes: AI becomes 'invisible' and ubiquitous; autonomous agents make business decisions; human skills become the differentiator; and AI infrastructure costs grow and start being charged.

• Prepare governance for agents (risk, auditability, accountability).
• Invest in human skills as competitive advantage.
• Plan AI FinOps/chargeback: inference and training costs hit the P&L.

View on LinkedIn

LinkedIn AI Agents

2025-12-29

Meta and Manus AI: agents with VMs, research, and deliverables

Note about Meta's investment/acquisition of Manus AI, an agent platform that spins up virtual machines to execute tasks, runs research, and generates deliverables (e.g., PPT decks). Based on hands-on use, it argues the backing can accelerate maturity and adoption.

• Agents with real execution (VMs) extend automation beyond chat.
• Good fits: research, asset generation, repeatable ops tasks.
• Set guardrails: permissions, reviews, and human-in-the-loop for critical actions.

View on LinkedIn

LinkedIn Automation

2025-10-06

OpenAI Agent Builder: is this the end of n8n or market expansion?

Response to OpenAI's Agent Builder and the 'panic' that automation tools (Zapier, n8n, Make) would die. The argument: a large share (~60% estimate) of automation is still traditional integrations and business logic, not AI. Agent Builder may expand the market and bring more visibility to low-code automation platforms.

• Separate deterministic automation from AI-driven, probabilistic steps.
• Agent Builder can be an entry point; low-code platforms stay essential.
• Audit workflows: where do you truly need AI vs pure integration?

View on LinkedIn

LinkedIn Decision Making

2025-09-26

Changing your mind is not weakness: integrity in decision-making

Leadership reflection on decision-making: stepping back is not weakness, but integrity when values or facts change. It highlights the courage to recognize an unhonorable commitment, take transparent responsibility, prioritize values over ego, and correct biased decisions. It ends with: 'stepping back is also a decision'.

• Design reversibility: decision records, checkpoints, kill switches.
• Normalize revision: changing your mind with data is maturity.
• Transparency and repair build trust more than stubborn promises.

View on LinkedIn

LinkedIn Prompt Engineering

2025-08-29

Prompt engineering: Role, Task, Audience, Tone, Constraints

Video/text primer on prompt engineering: vague prompts produce generic answers. It recommends giving instructions with five building blocks: Role, Task, Audience, Tone, and Format/Constraints. Emphasizes iteration: the first output is a draft; you are the director guiding the assistant.

• Role + Task align expertise and expected deliverable.
• Audience + Tone tailor depth and style (exec vs technical, etc).
• Constraints reduce verbosity and keep output within safe boundaries.

View on LinkedIn

LinkedIn NotebookLM

2025-08-28

NotebookLM: links as sources (YouTube/sites) and video summaries

NotebookLM update: beyond PDFs/text/audio, it now accepts website and YouTube links as sources, and can generate not only audio (podcast) summaries but also video summaries. The post demonstrates a test using four Context Engineering talks as inputs and requesting a video recap.

• Treat curated links as a mini knowledge base you can query.
• Video summaries speed up team alignment and stakeholder updates.
• Useful for research, onboarding, and literature review with traceable sources.

View on LinkedIn

LinkedIn Context Engineering

2025-08-23

Vibe coding + context engineering: beyond one-off prompts

Vibe Coding is co-programming with AI in an exploratory, fast flow state. The quality jump comes from Context Engineering: organizing objectives, constraints, examples, history, and references so the model understands the problem and stops being generic. It points to resources like YouTube, LangChain, BMAD Method, and Super Claude.

• Without context: more trial/error; with context: speed with quality.
• Standardize artifacts (living docs, story files) to keep project memory.
• Frameworks help operationalize context patterns in day-to-day dev.

View on LinkedIn

LinkedIn AI Adoption

2025-08-01

Generative AI with common sense: when classic RPA is better

Generative AI adoption with common sense: not everything should be automated with AI. Some work needs human judgment; in other cases, classic RPA can be faster and cheaper. The key is to decide where AI truly adds value versus where it only adds cost and complexity.

• Use a rubric: value, risk, cost, latency, explainability.
• Combine AI + RPA when useful (AI for classification, RPA for deterministic steps).
• Keep humans in the loop for sensitive decisions and external messaging.

View on LinkedIn

LinkedIn AI Agents

2025-07-24

Manus AI introduces itself: business + nerd in one agent

Quick experiment: asking Manus AI to introduce itself, highlighting a 'business and nerd' versatility that produces both relevant content and code. It mentions ongoing testing and a referral/code for extra credits.

• Use self-intro prompts to learn capabilities and limitations.
• Validate in two dimensions: executive outputs and technical outputs.
• Start with small real tasks before scaling adoption.

View on LinkedIn

LinkedIn Visuals

2025-07-19

Napkin AI: turn text into visuals in one click

Recommendation for Napkin AI (app.napkin.ai) to turn text into graphical visualizations quickly. Workflow: generate or upload content, select snippets, hit the 'lightning' button, and get many visual interpretations that summarize or expand the idea. Focus on productivity for communication and alignment.

• Great for turning drafts into diagrams without manual design.
• Use it for architecture, strategy, roadmaps, and pitch decks.
• Iterate fast: compare multiple visuals and pick the clearest one.

View on LinkedIn

LinkedIn Automation

2025-07-18

n8n with AI nodes: workflows that think before acting

How n8n levels up with native AI nodes: workflows can connect WhatsApp, databases, audio transcription, enrichment and summarization, and generate contextual responses. Automation shifts from pure if/then triggers to 'think before acting' steps (exceptions, writing, classification, translation), plugging OpenAI or Gemini directly into the flow without a separate backend.

• Architecture: low-code orchestration plus LLM steps inside the workflow.
• Use AI where ambiguity exists; keep the rest deterministic.
• Add governance: logs, prompt versioning, cost limits, and fallbacks.

View on LinkedIn

LinkedIn Security

2023-01-05

Flipper Zero: why bans can backfire and fuel clones

Take on Flipper Zero: similar legacy devices are easy to find on grey markets; Flipper's advantage is bundling functions and offering an intuitive UX to explore communication patterns with little prior study. The prediction is that bans can backfire, boosting visibility and encouraging clones.

• Bans rarely remove risk; focus on education and controls.
• Hardware accessibility will grow: prepare policies and detection.
• Defense should prioritize hardening and monitoring of radio/protocol surfaces.

View on LinkedIn