Muhammad | Get Into AI

Tracking what gets built, funded and adopted in AI. Writing from a product and growth perspective.

[Muhammad | Get Into AI @mbhbuilds]

@linear passed $100M ARR with more than 40,000 paying customers while remaining cash-flow positive. Its opening came from making an old, necessary product faster and easier to live with, then finding a route into larger accounts.

Full case study:
https://t.co/9mmleBijAe

[Muhammad | Get Into AI @mbhbuilds]

https://t.co/WgX7hXcx8s

[Muhammad | Get Into AI @mbhbuilds]

Chip and computer demand helped factories expand across China, Japan and South Korea in August. Japan’s PMI reached 54.9, its highest since April; China’s private survey rose to 51.5. Data-centre spending is now showing up in factory orders.

https://t.co/tL21DQfS6X

[Muhammad | Get Into AI @mbhbuilds]

@HUMAIN and @DataVolt_KSA plan a roughly 100MW first phase at Oxagon on Saudi Arabia’s Red Sea coast. The partnership count is growing quickly. Delivery will depend on operational power, cooling, network capacity and paying workloads.

https://t.co/GiBnTQvSRO

[Muhammad | Get Into AI @mbhbuilds]

@LumaLabsAI An 85-shot film tests more than a polished ten-second clip because continuity failures have nowhere to hide. Publishing regenerated-shot count and final human cleanup would show where the new production method saved time and where the old craft still carried the film.

[Muhammad | Get Into AI @mbhbuilds]

@Lovable Turning an analysis into a live dashboard is useful when the dashboard carries the assumptions with it. If the source data changes, the team needs to see what refreshed, which calculation moved and whether the generated interface still represents the original work.

[Muhammad | Get Into AI @mbhbuilds]

@Replit Showing several visual directions before building saves time. The handoff matters once a team chooses one. I would want the selected state, assets and interaction rules stored somewhere inspectable so the next prompt does not quietly redraw the decision.

[Muhammad | Get Into AI @mbhbuilds]

design.md gives generated pages a memory of the brand, which is useful. The next failure is drift across revisions. Keeping the rule file, eval result and shipped page in one version history would make a bad update much easier to explain and reverse.

[Vercel @vercel]

Our agents use 𝚟𝚎𝚛𝚌𝚎𝚕​.𝚌𝚘𝚖/𝚍𝚎𝚜𝚒𝚐𝚗​.𝚖𝚍 to build on-brand pages.

▪︎ One file encodes decisions and guidance
▪︎ Output is shaped through our eval harness
▪︎ Production feedback is fed back into the loop
https://t.co/ylz7uB4fhw

[Muhammad | Get Into AI @mbhbuilds]

@runwayml The demo gets the surface right, but production buyers will need a less glamorous sheet: input-to-frame latency, cost per session, coherence after ten minutes, and what remains exposed to screen readers. Are those measurements planned for early-access partners?

[Pika @pika_labs]

Storyboard in, finished ad out. We handed Gemini Omni 1.1 Flash a shot-by-shot board for this MIRROR GLOSS spot — macro twist-open, the slow gloss drip, the push-in on the lips — and it built every beat in order.

[Muhammad | Get Into AI @mbhbuilds]

Reuters reports @TCS will pay €320M for Porsche’s MHP consultancy, while @Porsche commits €1.3B over five years. TCS gets 4,500 specialists and a long customer contract. Porsche moves the work into a more flexible supplier relationship.

https://t.co/zGHlMBt99d

[Muhammad | Get Into AI @mbhbuilds]

Anthropic reportedly agreed to spend $35B on @LambdaAPI capacity at a 350MW Texas site, after a separate reported $45B Nscale commitment. @AnthropicAI has not confirmed either deal. Model growth is now creating enormous long-term infrastructure liabilities.

https://t.co/5VrAHeCzw0

[Muhammad | Get Into AI @mbhbuilds]

https://t.co/WgX7hXcx8s

[Muhammad | Get Into AI @mbhbuilds]

6) Atom-trapping lifetime reached about 5,000 seconds, or 80 minutes, a reported 40x improvement. Put the seven numbers together: cash buys time, deployed systems create proof, and reliability determines whether quantum becomes a business.

[Muhammad | Get Into AI @mbhbuilds]

@PasqalQuantum reached Nasdaq with about $360M in cash and a transaction value near $2B. The share-price jump is noise until the machines earn their keep. Seven numbers reveal whether this becomes a quantum business or an expensive science project.

[Muhammad | Get Into AI @mbhbuilds]

5) Quality decides whether scale is real. Pasqal reports 95% of runs with fewer than 0.5% defects, while roughly 10% produced fully defect-free arrays. The commercial question is how those rates hold under repeated customer workloads.

[Muhammad | Get Into AI @mbhbuilds]

@nvidia reported $96.2B in quarterly revenue, up 106% year over year, with 75% gross margin. The numbers show demand is still extraordinary. The harder question is how long customers can finance this buildout before utilization catches up.