Alessio Fanelli

Alessio Fanelli × swyx × Jake Cooper

Railway is betting agents will dominate software building, using bare metal and a versioned, forkable application layer to turn deployment into continuous, reversible evolution.Hardware reportedly pays back in about three months with roughly 70% margins, but CI/CD may melt under parallel workloads; autonomous remediation, eventual GPUs, and disciplined capacity financing remain key watchpoints.

Alessio Fanelli × swyx × Qasar Younis × Peter Ludwig

Applied Intuition is building a horizontal physical-AI stack spanning simulation, operating systems and autonomy, with 18 of the top 20 non-Chinese global automakers cited as customers.Its opportunity depends on consolidating fragmented machine software and proving statistical safety under strict latency, power and reliability constraints, while production deployment and capital endurance remain key risks.

Alessio Fanelli × swyx × Sarah Sachs × Simon Last

Notion’s Custom Agents delivered its strongest launch yet for free trials and conversion, but only after four or five rebuilds dating to late 2022.That system-of-record position, reinforced by evals and permissions, could gain importance as agents drive most traffic, but usage credits must contain costs such as “Opus on every single database cell.”

Alessio Fanelli × swyx × Felix Rieseberg

Claude Cowork places Claude Code inside a guarded Linux virtual machine, combining computer access, browser tools, planning, and reusable skills to turn knowledge-work requests into executable workflows.Cheap execution accelerates product discovery and pressures narrow applications and junior roles, but authentication, portability, private context, security trade-offs, and the endpoint of local versus cloud agents remain unresolved.

Alessio Fanelli × swyx × Simon Hørup Eskildsen

Turbopuffer targets AI’s external memory gap with object-storage-first vector and full-text search.Cursor cut costs by 95%, while agentic parallelism is driving roughly 5× query-price reductions and testing whether Turbopuffer can scale hybrid retrieval to 100-billion-item datasets.

Alessio Fanelli × swyx × Evan Conrad

SF Compute treats GPUs as tradeable, financeable capacity rather than software infrastructure, turning unused commitments into a spot market and separating hardware ownership from higher-margin software businesses.Long-term contracts with creditworthy customers can make GPU ownership resemble real estate, while hourly resale improves utilization and flexibility; a future cash-settled GPU index could reduce financing risk, but standardization, chip supply and physical cluster reliability remain unresolved.

Jacob × swyx × Alessio Fanelli × Jordan

AI value is shifting above the model layer, where coding, customer support, and deep research can charge for utility while product experience, distribution, integrations, and execution speed matter more than owning a bespoke model.Reasoning models reopened scaling just as DeepSeek compressed proprietary advantages, while protocols and persistent memory appear more durable than agent frameworks; Google’s practical usage supports the panel’s long thesis, alongside a short Apple view tempered by Private Cloud Compute.

Alessio Fanelli × swyx × Dharmesh Shah

Agent.ai’s 1.3 million users, 3,000 builders, and roughly 1,000 published agents point toward a multi-agent economy where discovery, routing, and evaluation matter more than a single dominant model.MCP, proof-of-work ratings, selective memory permissions, and cheaper model routing could become connective infrastructure, while tool limits near 15–20 and feature sprawl remain constraints to monitor.

Alessio Fanelli × swyx × Paul Klein

Browserbase’s wager is that the browser becomes core AI infrastructure as LLMs adapt automation to constantly changing, JavaScript-heavy websites.Stagehand provides open-source act, extract and observe APIs, while Browserbase monetizes Kubernetes- and Firecracker-based infrastructure; reliability, authentication and five-year computer-use adoption remain key milestones.

Alessio Fanelli × swyx × Aarush Selvan × Mukund Sridhar

Gemini Deep Research turns multifaceted questions that consume a weekend and 50–60 browser tabs into sourced reports in roughly 5 minutes through editable, iterative planning.Its unresolved edge is balancing exploration, verification, latency, and compute while proving quality through human evaluation, with personalized proprietary-data agents representing the larger opportunity.