You Can Only Achieve So Many 9s
Joined March 2007
- Tweets 5,281
- Following 222
- Followers 2,404
- Likes 1,991
203 Photos and videos
bfrench retweeted
Mar 2
94
220
2,764
1,012,516
When the world’s rate of change overtakes yours, “almost there” turns into a permanent condition. The horizon doesn’t just stay ahead — it actively escapes.
open.substack.com/pub/impert…
Jan 29
We've had a pretty wild new AI Agent feature that we've almost released twice, but then right as we get close, a new technical breakthrough makes the agent substantially better than before. The pace of new best practices emerging in AI is unlike anything that's happened in tech before.
The common trend is that we're all collectively figuring out the best ways to split work up between agents to enable them to do much more complicated tasks, handling vastly more information than the context window can natively support, finding the most relevant data to work with on the task, and mitigating the natural limits on the AI models.
What's amazing is that for many of these areas of improvements, you can simply decide to throw more compute at the problem at any point to unlock another tranche of performance improvement. Of course the ultimate constraint then is what someone will pay to perform the task, but it's amazing that we always have this variable at our disposal.
547
This is the Twin!
Jan 27
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗧𝘄𝗶𝗻 — 𝘁𝗵𝗲 𝗔𝗜 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗯𝘂𝗶𝗹𝗱𝗲𝗿.
No setup. Secure. Infinitely scalable.
We just raised a $𝟭𝟬𝗠 𝘀𝗲𝗲𝗱.
After a beta with 𝟭𝟬𝟬,𝟬𝟬𝟬 𝗮𝗴𝗲𝗻𝘁𝘀 𝗱𝗲𝗽𝗹𝗼𝘆𝗲𝗱, we’re now opening to everyone.
RT and comment “Twin” — first agents on us. 👇
276
Okay, let’s Twin up!
Jan 27
𝗜𝗻𝘁𝗿𝗼𝗱𝘂𝗰𝗶𝗻𝗴 𝗧𝘄𝗶𝗻 — 𝘁𝗵𝗲 𝗔𝗜 𝗰𝗼𝗺𝗽𝗮𝗻𝘆 𝗯𝘂𝗶𝗹𝗱𝗲𝗿.
No setup. Secure. Infinitely scalable.
We just raised a $𝟭𝟬𝗠 𝘀𝗲𝗲𝗱.
After a beta with 𝟭𝟬𝟬,𝟬𝟬𝟬 𝗮𝗴𝗲𝗻𝘁𝘀 𝗱𝗲𝗽𝗹𝗼𝘆𝗲𝗱, we’re now opening to everyone.
RT and comment “Twin” — first agents on us. 👇
257
The UI is the Friction: Eviction of the Apps is Intevitable
There is a specific kind of exhaustion reserved for the productivity enthusiast. It isn’t the tiredness of work done; it is the tiredness of work organized. I experience it often. I’m sure you know exactly what I’m talking about. Agents reduce fatigue when working with apps.
x.com/bfrench/status/2016152…
1
97
I wrote about this in 2023.
open.substack.com/pub/impert…
Jan 14
This is the inflection point for Gemini to overtake competitors.
All frontier models are basically even in performance for 99.9% of use cases.
The real value is in the integrations: how easily can I connect my data to AI.
Google has a massive competitive advantage here. And they're executing well against it.
Apple also has a competitive advantage here, and now with Gemini powering Siri, maybe they'll actually catch up.
133
I love clean air, but the environment doesn’t make the top ten for me.
Jan 4
A lot of Tesla haters and anti-EV folks start their argument with the assumption that people buy EVs because of the environment.
The truth is, I’m a conservative who drives a Tesla because it’s an insanely good product. The environmental benefits are just a bonus.
1
169
Visual Context Re-engineering.
complex creation (marketing strategies, deep research, video making) also requires having context visible and then iterating with feedback - canvas-style in flowith.
we visualize the 'rethought experience' to maximize your results.
3
165
AI “games” cannot be won. They can only be led.
And given recent experiments with @antigravity Chrome sub agents, Claude has fumbled the “game winning” touchdown run at the one.
31 Dec 2025
can't believe claude won the browser automation game with ... an extension 💀
2
165
AI productivity has many facets. If you're not using @typelessdotcom, you're moving through your daily work at a snail's pace.
Typing is outdated friction. Speaking is how ideas actually flow.
If you're drafting emails, notes, posts, or docs daily—Typeless turns thoughts into polished text effortlessly.
Try it free: typeless.com/?via=bill
It's on another level. 🔥
Typeless intelligently edits in real-time: removes fluff, fixes grammar, structures sentences, and makes it read like you carefully typed it. Natural flow, zero cleanup needed.
Context-aware magic: It adapts tone/style based on the app (professional emails vs. casual chats) and understands intent. Speak mixed languages? No problem—it detects and handles seamlessly. Alternatives often choke on accents, code-switching, or nuance.
Truly hands-free & 4x faster: Whisper mode for quiet spots, speak-to-edit (just say 'change that to...'), and it works everywhere—no app-switching.
Most competitors are clunky, inaccurate in noise, or require constant corrections. Privacy-first (no creepy data hoarding): Everything stays on-device. Zero retention or training on your voice/data. Big players? They often mine your speech for ads/training.
Productivity gamechanger!
1
2
103
If you think only Claude can run skills, look closer at Gemini.
>>> most mcp setups today just load a massive pile of tool definitions into context upfront.
This is not entirely true. If you examine how Gemini works, it discretely rejects portions of your context, optimizing the window AND all while loading only the model parameters needed to perform the inference. It is a true MOE architecture that no other frontier models have.
It allows the model to have the "intelligence" of a giant model with the "speed" of a much smaller one.
As it pertains to saturated skill contexts submitted in the prompt of via MCP, it does not "load" the unnecessary skills in terms of active computation.
Because Gemini uses a Mixture of Experts (MoE) architecture, it treats "skills" and "subskills" with extreme surgical precision. Even if your prompt contains a massive context that covers a broad skill, the model will only use the specific active parameters required to process the exact token it is currently generating.
Google’s Gemini PRO: Next-generation AI model | Encord encord.com/blog/google-gemin…
I have proven this to myself using Claude Skills. It actually runs Claude Skills better than Claude itself.
146
Yep. It exploded for me in early 2025 with @typelessdotcom.
15 Dec 2025
Voice AI is going to explode in 2026. Here’s what I’m seeing:
1. Dictation has completely changed how I work
I go on walks where I dictate to Otter for 40min. I built an app this weekend while lifting weights. The productivity gain is real.
2. Phone booths everywhere
I visited the offices of two large AI companies last week. They have phone booths everywhere. I watched someone walk into one, dictate, then walk right back out. That’s it.
3. Microphones at every desk
I visited Wispr Flow’s headquarters in October (see pic). Every single employee had a $60 microphone on their desk and can whisper tasks all day long to AI. You cannot hear them even if you’re at the neighboring table.
4. OpenAI says typing is the bottleneck
Alexander Embiricos, head of product for Codex, just went on the @lennysan podcast that the “current underappreciated limiting factor” to AGI-level productivity isn’t model capability, it’s human typing speed.
We are literally being held back by our fingers.
113