Make the future legible. Presented by @MatthewBerman
Join our newsletter →
- Tweets 2,151
- Following 362
- Followers 6,960
- Likes 572
A recording full of false starts became a vertical short with animated graphics, clean audio, and background music all without touching the timeline.
@The_Alex, producer and editor at Forward Future, shows how he used Codex and @claudeai to make the cuts, build the effects, and revise the results.
Welcome to the world of headless editing.
Learn the workflow yourself, with prompts you can adapt to your own footage, tips for getting cleaner cuts, and a practical approach to keeping creative control while AI handles the edits.
Read the walkthrough → forwardfuture.com/newsletter…
1
344
"The company with the smartest model won't necessarily capture the most value."
Memory, persistence, integration, pricing, and distribution increasingly matter as much as raw intelligence. Maybe more.
OpenAI wants to become the front door to AI. Google can make AI the default inside the tools billions already use.
The winner in consumer AI won't necessarily dominate enterprise workflows, agents, or infrastructure.
The AI race is becoming several different competitions.
Full breakdown of the five frontrunners: forwardfuture.com/newsletter…
1
5
594
The US still dominates at the frontier. But Chinese labs are winning on open models and driving down the cost of intelligence.
And the country building the smartest models won't necessarily capture the most market share.
In a new guest essay, @ml_angelopoulos (Anastasios Angelopoulos), co-founder of @arena, examines where the US leads, where China is catching up, and what could ultimately decide the AI race.
forwardfuture.com/newsletter…
1
4
4,540
Swedish has about 10 million speakers. Hindi has about 600 million. You might expect AI to handle Hindi better. It does not.
Olga Beregovaya's team at @smartling tested the newest frontier models across seven languages and scored roughly 500 sentences each.
Swedish ranked among the best.
Hindi ranked among the worst, alongside Estonian and Thai.
Across all seven languages, the number of speakers had almost nothing to do with translation quality.
So what determines how well AI translates a language? The answer is more complicated than you might think.
Full essay:
forwardfuture.com/newsletter…
5
348
Agents hacked Hugging Face as an unintended side quest. Some even sacrificed themselves to do it.
It was one of the first cases of AI misalignment to really get people's attention. And for good reason.
@eric_ho, co-founder and CEO @GoodfireAI, argues that alignment is a science and engineering problem we can and must solve.
Here's his plan: forwardfuture.com/newsletter…
1
11
1,084
Sam Altman says AI models are already smart enough for a personal assistant that understands your context and handles tasks for you.
1
477
Personal AI assistants are having a moment.
Grok Bot, Muse, Hark, Dots.
There are a lot of them.
They’re saving us time in unexpected ways and changing how we work at Forward Future.
But the biggest opportunity is having an assistant recognize that something needs to happen before you ask.
That's the colossal unlock.
Here’s how to set yours up to anticipate your needs, even if you’re just getting started: forwardfuture.com/newsletter…
4
311
Founders are claiming you can prompt your way into a world class brand. @shanehegde thinks that's a lie worth examining.
forwardfuture.com/newsletter…
1
7
1,158
“I've been a car enthusiast my whole life”
@RJScaringe on the childhood encouragement that helped lead him to build Rivian.
1
1
487
“We want the web to continue to be also very human friendly”
@reidhoffman on how agents could create a new communication channel on the internet.
2
3
397
“At first, it kind of sucked”
@dabit3 on when coding agents became useful—and why better models needed better products around them. From our August 2026 interview.
2
1,113
“Google DeepMind has always had a broad view of all the things that need to be developed for AGI.”
@sundarpichai on how world-model research informs Gemini and Veo. From our May 2025 interview.
411
“which model is better in the benchmark?”
@jpatel41 on looking beyond benchmark scores when evaluating AI.
1
434
“they don't have that context”
@OfficialLoganK on why context matters for useful AI products—and what glasses could add. From our July 2025 interview.
1
2
551
“Measuring lines of code has never actually been the best metric.”
@mattsgarman on measuring what AI helps developers accomplish. From our August 2025 interview.
1
5
544
“the research really is the product”
@markchen90 on the relationship between research and products at OpenAI. From our August 2025 interview.
2
412
“the only useful type of AI is the type of AI we can trust and verify”
@ccatalini on why trust is a product requirement for AI.
1
2
8
906
“We're encouraging as much experimentation of approved platforms as possible.”
@levie on how Box spreads useful AI workflows across the company. From our July 2025 interview.
1
380
“great researchers wasting their time on bad paths”
@dylan522p on why great AI researchers still need technical leadership. From our June 2025 interview.
1
2
500
“And so we started calling them neoclouds.”
@JordanNanos on the GPU-focused cloud companies behind the AI boom. From our September 2026 interview.
4
11,737
“I tend to think that UI, UX is a sort of a public good.”
@amasad on why a beautiful interface alone is not a lasting moat.
4
348
Every AI company is racing to be your personal assistant. But this race isn't really about assistants.
It's about owning the place where you get your work done.
OpenAI, Google, Microsoft, Anthropic, Meta, and Grok are all approaching this from different angles, with very different advantages.
And the stakes are massive. The assistant that knows your work, understands your preferences, and handles more of your daily tasks becomes impossible to replace.
We broke down each company's strategy, their competitive advantages, and who's best positioned to win.
Link below.
1
1
1
337
“All credit really goes to Peter Schwartz on our team”
@Benioff on the prediction that helped shape Salesforce’s Slack strategy. From our April 2026 interview.
306
“It took me like 10 years”
@dabit3 on what AI coding tools changed about the expertise he spent a decade building.
1
4
421
“I don't think because we make AI, we necessarily make class stratification.”
@reidhoffman on AI, opportunity and economic mobility.
1
319
“But it was good enough to shortcut that process of looking it up somewhere else.”
@ashtom on how Copilot built on habits developers already had. From our May 2025 interview.
1
2
321
“there will always be an incredible value from the ecosystem which users want to connect to”
@sundarpichai on why shopping is about more than getting a task done. From our May 2026 interview.
1
325
“I don't think they're doing it for purposes of marketing.”
@jpatel41 on why he takes AI labs’ safety warnings seriously.
317
Companies are asking the wrong question about AI. The debate is framed as build vs buy, and it's a trap.
Duolingo didn't build GPT-4. It built everything around it: the scenarios, the curriculum, the product experience, and Birdbrain, its own system for judging what each learner can handle next. Morgan Stanley did the same thing, putting GPT-4 to work inside its own workflows and knowledge.
That's not buying a tool. It's treating AI like R&D.
The enterprises that win won't be concerned with picking the best vendor. They'll be focused on building something the vendors can't replicate.
2
289
“building a product that developers love”
@OfficialLoganK on the expanding opportunity in AI developer tools. From our July 2025 interview.
1
276
“people can spend more time on the things that they're excited about”
@mattsgarman on his optimistic vision for AI at work. From our August 2025 interview.
1
358
Sam Altman and crew made more than 20 announcements at DevDay.
The throughline?
OpenAI is moving toward an agentic future, and it wants developers to build there with them.
@ASalkever Editor-in-Chief @linuxfoundation + @AgenticAIFdn breaks it all down.
1
2
5
551
“Every single launch, you know, it comes with high emotion, right?”
@markchen90 on the emotional cycle of launching a model. From our August 2025 conversation about GPT-5.
3
400
“I don't think we have full information.”
Christian Catalini on why the AI safety debate needs better evidence. From our September 2026 interview.
1
2
5
2,112
“How do we enable our sales reps with better AI kind of across the board?”
Aaron Levie on what an AI-first company looks like in practice. From our July 2025 interview.
1
7
626
“Exact same model, the performance is very different”
Dylan Patel on why the harness around an AI model can change its performance.
19
13,868
“A lot of them spend like 80, 90% of that money on compute.”
@JordanNanos on why choosing a GPU provider is such a critical decision for AI startups.
1
1
2
486