English
Based on โLIVE VIBE CHECK: GPT-5.5 Has it allโ from Every Watch the original video
OpenAIโs Silent Revolution: GPT-5.5 Emerges as the AI Workhorse You Didnโt See Coming
A new titan has entered the artificial intelligence arena, and according to a panel of seasoned AI experts, itโs not just a contender โ itโs a game-changer. GPT-5.5, OpenAIโs latest model, has been put through its paces in a rigorous three-week โvibe check,โ revealing a surprising blend of senior engineering prowess and everyday utility that has some calling it their new โdaily driver.โ While not yet available on the API, its rollout to Codex and ChatGPT promises to reshape how developers and knowledge workers interact with AI.
The consensus from the Every.to team, a subscription service dedicated to staying at the forefront of AI, is clear: GPT-5.5 is โreally great,โ โsuper collaborative,โ โsuper fast,โ and even โpretty personable.โ But its true distinction lies in a rare duality โ its ability to excel at highly complex, senior-level engineering tasks while simultaneously serving as an indispensable workhorse for a broad spectrum of daily activities.
The Workhorse Unveiled: Benchmarking a New Era
Dan, the host and primary tester, unveiled the headline finding: โGPT-5.5 is OpenAIโs workhorse model.โ What makes it so? The teamโs proprietary senior engineer benchmark, designed to evaluate a modelโs capacity to rewrite an existing codebase with the insight and skill of an experienced human engineer, delivered astonishing results. Tested against a real codebase, rewritten independently by two senior engineers, GPT-5.5 achieved a best score of 62 out of 100. In stark contrast, Opus 47, a highly regarded competitor, scored a mere 33. This nearly 30-point swing signifies a monumental leap in AIโs ability to tackle sophisticated code refactoring.
However, a fascinating caveat emerged: GPT-5.5 achieved its peak performance when guided by a plan literally written by Opus 47. This intricate interplay suggests a compelling division of labor: Opus excels at high-level strategic planning, offering โterse, spec-like, and contract-heavyโ blueprints (e.g., โtake this gigantic file and get it down to 500 linesโ), while GPT-5.5 demonstrates an unparalleled capacity for relentless, assertive execution.
Dan elaborated on this dynamic: โIf you give it a prompt that says, โHey, I want you to like go and rewrite a major part of this codebaseโฆ figure out how you would rewrite it from first principles and then do it,โ both 4.7 and 5.5 can figure outโฆ a plan for it.โ But when it comes to execution, Opus 47 often โgets distracted and a little bit almost intimidated by a really big codebase and a big rewrite,โ opting for minor patches. GPT-5.5, on the other hand, possesses the โcourage or assertivenessโ to โdelete a bunch of code andโฆ think about this from first principles without getting as distracted by the existing code.โ This tenacious execution, sustained over โmany turns, over many many many hours, over many many tokens,โ is a novel capability not seen in previous models.
A Model for Every Task? Diverse Perspectives from the Front Lines
The teamโs diverse roles and workflows offered a multifaceted โvibe check,โ revealing how GPT-5.5 adapts to different professional needs.
Reliability for the Enterprise: Mike Taylorโs Perspective
Mike Taylor, Everyโs Head of AI Consulting, found GPT-5.5 to be โthe most reliable model I tested.โ Comparing it to a โsafe Waymoโ versus Opusโs โdangerous Tesla,โ Mike highlighted its dependability for tasks where he couldnโt afford to micromanage. For instance, creating curriculum for corporate training materials, which involves sifting through โtons of call notes from different people across the organizationโ to ensure all themes and issues are represented, became a seamless process.
โIt just like hasnโt failed on that once,โ Mike stated, contrasting it with Opus, which, despite producing โreally sharp stuffโ and โcool titles,โ often required line-by-line review. For corporate training, โyou donโt want cool titles all the timeโฆ you actually want something dependable, reliable that like, you know, regular people will find accessible.โ GPT-5.5 proved ideal for tasks needing to be โunoffensive,โ โreliable,โ and comprehensive without being โa little bit wild.โ
The Generalist vs. The Specialist: Kieran Classenโs Nuanced View
Kieran Classen, GM of Cora and creator of Compound Engineering, offered a more nuanced take. While acknowledging GPT-5.5โs impressive coding capabilitiesโscoring identically to Opus 4.7 on his specific coding benchmarkโhe views it as more of a โspecialistโ than a โgeneralist.โ
โClaude is the generalist that is a very good coding model, but itโs also very good at product workโฆ looking at the big picture,โ Kieran explained. GPT-5.5, conversely, โis very good in execution and going into details but sometimes it like breaks down if you look at it from far away and you just see things not being coherent.โ As a โproduct engineer generalistโ building a new version of Cora, which involves wide-ranging work from front-end to back-end, he still reaches for Opus 4.7 as his daily driver.
Kieran also noted a significant limitation: its performance with Ruby. While โvery, very good atโฆ React, like a Next appโ and TypeScript, he found it โjust not good at Ruby,โ which is Coraโs primary language. โThatโs not how you write Ruby,โ he lamented, making it a โbiggest blockerโ for his specific workflow.
The Vibe Coderโs Dream: Naveenโs โWell-Roundedโ Experience
Naveen, GM of Monologue, represents the โengineer engineerโ perspective, aligning more with OpenAIโs approach. He describes GPT-5.5 as a โreally well-rounded model,โ a significant shift from his previous preference for Claude/Opus for coding. He now uses 5.5 for everything from Python and Swift codebases to support replies for Monologue.
Naveenโs most compelling example is his โvibe codingโ adventures. While recovering from pink eye, he leveraged 5.5 to create three different apps from scratch, including โDayline,โ a Raycast alternative for daily to-dos. He simply gave it a screenshot of Raycast and a high-level idea, and 5.5 โvibe codedโ the entire Mac app, including complex minor interactions, in a single thread spanning โ200 million tokens.โ He even expanded it to an iOS app that syncs automatically.
โI didnโt look at single code,โ Naveen emphasized. This incredible feat highlights 5.5โs ability to maintain context over extremely long conversations and manage multiple codebases simultaneously, a testament to its robust โcompactionโ or context management. Mike Taylor echoed this, noting that his own โKarpathy-style knowledge baseโ project, initially needing a โRalph Wiggum loopโ (a constant human-like check-and-commit process), now runs faster with 5.5 simply compacting itself, requiring โless harness essentially.โ
Speed as a Superpower: Austinโs Shift in Knowledge Work
Austin, Everyโs Head of Growth, presented perhaps the most dramatic shift in allegiance. Previously a dedicated Claude Code user, he now finds the Codex app, powered by GPT-5.5, to be his โdaily driver for everything I do.โ For someone without a technical background, 5.5 has made engineering work, from building dashboards to shipping landing pages and creating strategic plans, โway more comfortable.โ
A few months prior, Austin found older Codex models alienating, making him โfeel like Iโm dumb.โ The models would ask clarifying questions in a tone that felt dismissive. โAll of that is gone for me in the new model,โ he stated. โI both understand what itโs saying [and] I also really trust what itโs saying.โ
The speed of the Codex app with 5.5 is a crucial factor. Austin described pointing the model at Notion and Slack, brain-dumping high-level campaign goals, and receiving a campaign plan that was โ90% of what it came up with.โ This efficiency allows him to manage complex tasks โwhile Iโm in meetingsโ or โwatching the NBA playoffs,โ nudging the AI along with only โ10% of my brain working.โ This โspeed is in some waysโฆ a type of intelligence,โ Dan added, granting users โa lot more power.โ
Creative Crossroads: Design and Imagery
While GPT-5.5 excels in many areas, the panel noted mixed results in creative tasks, particularly design. Kieran observed that while typography and structural alignment in 5.5โs designs looked better, the overall aesthetic could be โa little bit chaotic,โ sometimes showing โsome degradationโ compared to previous models. He found Opus 4.7 still performed better on subjective โcozy islandโ design tests. Austin also found Opus models superior for video clipping and image generation via tools like ThreeMotion, noting 5.5โs โlaziest possible approachโ of simply zooming in on recordings.
However, a new contender has emerged from OpenAI itself: the new GPT image generation tool. Both Dan and Austin were โblown awayโ by its capabilities. Austin, initially hesitant, used it to create the YouTube thumbnail for their very live stream in โlike 3 minutes,โ requiring โno notes.โ It produced high-quality, clean icons, accurate finger counts (without a source image), and even a โ95% approximation of our logo.โ Dan added that itโs โvery good at like, โOh, I want you to change this one little thing, but keep the rest the same,โโ and impressively, it can generate accurate likenesses of people. This suggests OpenAI is rapidly closing the gap in creative visual domains.
The Evolving AI Landscape: A Luxury of Choice
The โvibe checkโ on GPT-5.5 paints a picture of an AI landscape rich with specialized and increasingly powerful models. While GPT-5.5 distinguishes itself as an assertive, high-performance workhorse, particularly in complex engineering execution and fast-paced knowledge work, other models like Opus 4.7 retain their edge in strategic planning and certain creative design tasks.
The expertsโ varied experiences underscore a crucial point: โAll the models are very goodโฆ Thereโs nothing bad about any of these models. Theyโre amazing.โ The choice now often boils down to subtle nuances, specific workflow alignment, and even the language being used. OpenAIโs GPT-5.5 has undoubtedly raised the bar, offering a glimpse into a future where AI not only assists but actively drives complex projects with unprecedented speed and reliability, empowering users to achieve more with less friction. The revolution, it seems, is well underway, and itโs happening at warp speed.
ํ๊ตญ์ด
โLIVE VIBE CHECK: GPT-5.5 Has it allโ โ Every ๊ธฐ๋ฐ ๊ธฐ์ฌ ์๋ณธ ์์ ๋ณด๊ธฐ
GPT-5.5: OpenAI์ ์๋ก์ด ์ฃผ๋ ฅ ๋ชจ๋ธ, ๊ฐ๋ฐ๋ถํฐ ์ง์ ์ ๋ฌด๊น์ง โ์ฌ๋ผ์ด๋โ์ ํ์
์ธ๊ณต์ง๋ฅ(AI) ๊ธฐ์ ์ ์ต์ ์ ์์ ๋์์์ด ์๋ก์ด ๋ชจ๋ธ์ ํ์ํ๋ Every ํ์ด ์ต๊ทผ ์ถ์๋ OpenAI์ GPT-5.5 ๋ชจ๋ธ์ ๋ํ ์ฌ์ธต์ ์ธ โ๋ฐ์ด๋ธ ์ฒดํฌ(Vibe Check)โ ๊ฒฐ๊ณผ๋ฅผ ๊ณต๊ฐํ์ต๋๋ค. ์ง๋ 3์ฃผ๊ฐ์ ์ง์ค์ ์ธ ํ ์คํธ๋ฅผ ํตํด, GPT-5.5๋ โ์ ๋ง ๋๋จํ๋ค(really fucking great)โ๋ ํ๊ฐ์ ํจ๊ป ๊ฐ๋ฐ์๋ถํฐ ๋ง์ผํฐ์ ์ด๋ฅด๊ธฐ๊น์ง ๋ค์ํ ์ ๋ฌธ๊ฐ๋ค์ ์ผ์ ์ ๋ฌด๋ฅผ ํ์ ํ ์ ์ฌ๋ ฅ์ ์ ์ฆํ์ต๋๋ค. ์ด ๊ธฐ์ฌ๋ Every ํ์ ๋ค๊ฐ์ ์ธ ๊ด์ ์ ํตํด GPT-5.5์ ํต์ฌ ์ญ๋, ๊ฐ์ , ๊ทธ๋ฆฌ๊ณ ์ฌ์ ํ ์กด์ฌํ๋ ๋ฏธ๋ฌํ ์ฐจ์ด์ ์ ๋ถ์ํฉ๋๋ค.
GPT-5.5: OpenAI์ ์๋ก์ด ์ฃผ๋ ฅ ๋ชจ๋ธ
Every ํ์ CEO ๋(Dan)์ GPT-5.5๊ฐ ์ด๋ฏธ ๊ทธ์ โ๋ฐ์ผ๋ฆฌ ๋๋ผ์ด๋ฒ(Daily Driver)โ๊ฐ ๋์๋ค๊ณ ์ ์ธํ๋ฉฐ, ์ด ๋ชจ๋ธ์ด ์ด์ GPT-5์ ๋ฏธ์ธ ์กฐ์ ๋ฒ์ ์ด ์๋, โ์คํผ๋(Spud)โ ๋ชจ๋ธ๋ก ์๋ ค์ง ์์ ํ ์๋ก์ด ์ฌ์ ํ์ต ๋ชจ๋ธ์์ ๊ฐ์กฐํ์ต๋๋ค. ํ์ฌ API๋ก๋ ์ถ์๋์ง ์์์ง๋ง, ์ด๋ฏธ ChatGPT์ Codex๋ฅผ ํตํด ์ฌ์ฉ์๋ค์๊ฒ ๋ฐฐํฌ๋๊ธฐ ์์ํ์ผ๋ฉฐ, ๊ทธ ๊ฐ๋ ฅํจ ๋๋ฌธ์ OpenAI๊ฐ ์ ์คํ ํ ์คํธ๋ฅผ ๊ฑฐ์น๊ณ ์๋ค๊ณ ์ธ๊ธํ์ต๋๋ค.
์๋์ด ์์ง๋์ด ๋ฒค์น๋งํฌ์ ์๋์ ์ฑ๊ณผ
Every ํ์ ๋ชจ๋ธ์ด ๊ธฐ์กด ์ฝ๋๋ฒ ์ด์ค๋ฅผ ์๋์ด ์์ง๋์ด์ฒ๋ผ ์ฌ์์ฑํ๋ ๋ฅ๋ ฅ์ ํ๊ฐํ๋ ๋ ์์ ์ธ โ์๋์ด ์์ง๋์ด ๋ฒค์น๋งํฌโ๋ฅผ ์ด์ํฉ๋๋ค. GPT-5.5๋ ์ด ๋ฒค์น๋งํฌ์์ 100์ ๋ง์ ์ 62์ ์ ๊ธฐ๋กํ๋ฉฐ, ๊ฒฝ์ ๋ชจ๋ธ์ธ Opus 4.7์ ์ต๊ณ ์ ์ 33์ ๊ณผ ๋น๊ตํ์ ๋ ์ฝ 30์ ๊ฐ๊น์ด ๋์ ์ ์๋ฅผ ๋ฐ์์ต๋๋ค. ์ด๋ GPT-5.5๊ฐ ๋ณต์กํ ์์ง๋์ด๋ง ์์ , ํนํ ์ฝ๋ ์ฌ์์ฑ ๋ฅ๋ ฅ์์ ํ์ํจ์ ๋ณด์ฌ์ค๋ค๋ ์ฆ๊ฑฐ์ ๋๋ค.
ํฅ๋ฏธ๋ก์ด ์ ์ GPT-5.5๊ฐ ๊ฐ์ฅ ๋์ ์ ์๋ฅผ ๊ธฐ๋กํ ๋, Opus 4.7์ด ์์ฑํ โ๊ณํ(plan)โ์ ์ฌ์ฉํ๋ค๋ ๊ฒ์ ๋๋ค. Opus 4.7์ ๊ณํ์ ๊ฐ๊ฒฐํ๊ณ ๋ช ์ธ์ ์ด๋ฉฐ โ๊ฑฐ๋ํ ํ์ผ์ 500์ค๋ก ์ค์ฌ์ผ ํ๋คโ์ ๊ฐ์ ๊ตฌ์ฒด์ ์ธ ๋ชฉํ๋ฅผ ์ ์ํฉ๋๋ค. Opus 4.7 ์์ฒด๋ ์ด๋ฌํ ๊ฑฐ๋ํ ์ฌ์์ ํ๋ก์ ํธ๋ฅผ ์คํํ๋ ๋ฐ ์ฃผ์ ํ๊ฑฐ๋ ์์ ๋ถ๋ถ๋ง ์์ ํ๋ ค ํ์ง๋ง, GPT-5.5๋ ์ด ๊ณํ์ ๋ฐ์๋ค์ฌ ์๋ง์ ํด๊ณผ ํ ํฐ์ ๊ฑธ์ณ ์คํํ๊ณ , ์ฌ์ง์ด ๋ง์ ์์ ์ฝ๋๋ฅผ ์ญ์ ํ๋ โ์ฉ๊ธฐ(courage)โ์ โ๋จํธํจ(assertiveness)โ์ ๋ณด์ฌ์ค๋๋ค. ์ด๋ GPT-5.5๊ฐ ๋๊ท๋ชจ ์ฝ๋๋ฒ ์ด์ค๋ฅผ ์ฒซ ๋ฒ์งธ ์์น๋ถํฐ ์ฌ์์ฑํ๋ ๋ฐ ์์ด ๋ ๋ณด์ ์ธ ๋ฅ๋ ฅ์ ๊ฐ์ก์์ ์์ฌํฉ๋๋ค.
๋ค์ํ ์ฌ์ฉ์ ๊ฒฝํ: โ๋ฐ์ผ๋ฆฌ ๋๋ผ์ด๋ฒโ๋ถํฐ โ์ ๋ฌธ๊ฐโ๊น์ง
Every ํ์ ๋ค์ํ ์ ๋ฌธ๊ฐ๋ค์ GPT-5.5์ ๋ํด ๊ฐ๊ธฐ ๋ค๋ฅธ ๊ฒฝํ๊ณผ ๊ด์ ์ ๊ณต์ ํ๋ฉฐ, ์ด ๋ชจ๋ธ์ ๋ค๋ฉด์ ์ธ ํน์ฑ์ ๋๋ฌ๋์ต๋๋ค.
๋ง์ดํฌ ํ ์ผ๋ฌ: ์ ๋ขฐํ ์ ์๋ โ์์ ํโ ์กฐ๋ ฅ์
Every์ AI ๊ธฐ์ ์ปจ์คํ ์ฑ ์์์ธ ๋ง์ดํฌ ํ ์ผ๋ฌ(Mike Taylor)๋ GPT-5.5๋ฅผ โ๊ฐ์ฅ ์ ๋ขฐํ ์ ์๋ ๋ชจ๋ธโ์ด๋ผ๊ณ ํ๊ฐํ์ต๋๋ค. ๊ทธ๋ Opus๋ฅผ โํ ์ฌ๋ผ(Tesla)โ์ฒ๋ผ ํฅ๋ฏธ๋กญ์ง๋ง ๋ค์ ์ํํ ๋ชจ๋ธ๋ก ๋น์ ํ๋ฉฐ, GPT-5.5๋ โ์จ์ด๋ชจ(Waymo)โ์ฒ๋ผ ํธ์ํ๊ณ ์์ ํ๊ฒ ๋๊ปด์ง๋ค๊ณ ๋งํ์ต๋๋ค. ๊ทธ๋ ์ธ์ฌํ ๊ด๋ฆฌ๊ฐ ํ์ํ ์์ ์๋ Opus๋ฅผ ์ ํธํ์ง๋ง, ํฌ๊ฒ ์ ๊ฒฝ ์ฐ์ง ์์๋ ์์ ํ๊ฒ ์๋ฃ๋ ๊ฒ์ด๋ผ๊ณ ๋ฏฟ๋ ์์ (์: ๊ต์ก ์๋ฃ ์ปค๋ฆฌํ๋ผ ์์ฑ)์๋ GPT-5.5๋ฅผ ์ฌ์ฉํฉ๋๋ค. ํนํ, ๊ธฐ์ ๊ต์ก๊ณผ ๊ฐ์ด โ๋ถ์พํ์ง ์๊ณ (unoffensive)โ, โ์ ๋ขฐํ ์ ์์ผ๋ฉฐ(reliable)โ, โ์ ๊ทผ์ฑ ์๋(accessible)โ ๊ฒฐ๊ณผ๋ฌผ์ ์๊ตฌํ๋ ๊ฒฝ์ฐ GPT-5.5๊ฐ ํ์ํ๋ค๊ณ ๊ฐ์กฐํ์ต๋๋ค. ์ด์ Opus๋ โ๋ฉ์ง์ง๋ง ๋ค์ ๊ฑฐ์น(wild)โ ๊ฒฐ๊ณผ๋ฌผ์ ๋ด๋์ ์ผ์ผ์ด ์์ ํด์ผ ํ๋ ๋ฒ๊ฑฐ๋ก์์ด ์์๊ธฐ ๋๋ฌธ์ ๋๋ค.
ํค๋ฐ ํด๋์จ: ๊ฐ๋ ฅํ โ์ ๋ฌธ๊ฐโ ๋ชจ๋ธ, ๊ทธ๋ฌ๋ ๋ฃจ๋น๋ ๊ธ์์
Cora์ GM์ด์ Compound Engineering์ ์ฐฝ์์์ธ ํค๋ฐ ํด๋์จ(Kieran Classen)์ GPT-5.5๋ฅผ โ๋งค์ฐ ํ๋ฅญํ ๋ชจ๋ธโ๋ก ์ธ์ ํ๋ฉด์๋, ๊ทธ๋ฅผ ์ํ โ๋ฐ์ผ๋ฆฌ ๋๋ผ์ด๋ฒโ๋ ์๋๋ผ๊ณ ๋งํ์ต๋๋ค. ๊ทธ๊ฐ ์ํํ๋ ์ ํ ๊ฐ๋ฐ(ํ๋ฐํธ์๋, ๋ฐฑ์๋, ํ ์คํธ ๋ฑ)์ ๊ด๋ฒ์ํ ์ ๋๋ด๋ฆฌ์คํธ์ ์ญ๋์ ์๊ตฌํ๋๋ฐ, ๊ทธ๋ GPT-5.5๊ฐ โ์ ๋๋ด๋ฆฌ์คํธโ๋ณด๋ค๋ ํน์ ์์ญ์ ํนํ๋ โ์ ๋ฌธ๊ฐ(specialist)โ์ฒ๋ผ ๋๊ปด์ง๋ค๊ณ ์ค๋ช ํ์ต๋๋ค.
๊ทธ๋ Claude๋ฅผ ์ ํ ์์ , ํฐ ๊ทธ๋ฆผ ๋ณด๊ธฐ, ์ธ๋ถ ์ฌํญ ํ์ ๋ฑ์์ ๋ฐ์ด๋ ์ ๋๋ด๋ฆฌ์คํธ๋ก ๋ณด๋ ๋ฐ๋ฉด, GPT-5.5๋ ์คํ๋ ฅ๊ณผ ์ธ๋ถ ์ฌํญ ์ฒ๋ฆฌ์๋ ๋งค์ฐ ๊ฐํ์ง๋ง, ๋๋ก๋ ์ ์ฒด์ ์ธ ์ผ๊ด์ฑ์ด ๋ถ์กฑํ ๋๊ฐ ์๋ค๊ณ ์ง์ ํ์ต๋๋ค. ํนํ, ๊ทธ๊ฐ ์ฃผ๋ก ์ฌ์ฉํ๋ ํ๋ก๊ทธ๋๋ฐ ์ธ์ด์ธ Ruby์ ๋ํด์๋ GPT-5.5๊ฐ โ์ข์ง ์๋ค(not good at Ruby)โ๊ณ ๋งํ๋ฉฐ, TypeScript์ ๊ฐ์ด OpenAI๊ฐ ์ฃผ๋ก ํ์ตํ ์ธ์ด์์๋ ๋๋ผ์ด ์ฑ๋ฅ์ ๋ณด์ด์ง๋ง, Ruby์์๋ ๊ทธ ๋ฅ๋ ฅ์ด ํ์ ํ ๋จ์ด์ง๋ค๊ณ ๊ฐ์กฐํ์ต๋๋ค.
ํ์ง๋ง GPT-5.5์ ๊ฐ๋ ฅํจ์ ๋ถ์ธํ ์ ์์ต๋๋ค. ํค๋ฐ์ ๋จ ํ๋์ ํ๋กฌํํธ๋ก ๋ง์ถคํ ๊ณ ๋ฌด ์ค๋ฆฌ ์์ ์ ๋ง๋๋ ์น ์ฑ(React, Next.js)์ โ์์ท(one-shot)โ์ผ๋ก ๊ตฌํํ๋ ๋ฐ ์ฑ๊ณตํ๋ค๊ณ ๋ฐํ์ต๋๋ค. ์ด๋ ์ด์ Opus 4.7์์๋ ๋ถ๊ฐ๋ฅํ๋ ๊ฒฝํ์ ๋๋ค. ๋์์ธ ์ธก๋ฉด์์๋ ํ์ดํฌ๊ทธ๋ํผ์ ๊ตฌ์กฐ๋ ๊ฐ์ ๋์์ง๋ง, ์ ๋ฐ์ ์ธ ์๊ฐ์ ๋์์ธ์ ์ด์ ๋ชจ๋ธ์ด๋ Opus๋ณด๋ค ๋ค์ โํผ๋์ค๋ฝ๊ฑฐ๋(chaotic)โ โ์ด์ํ(weird)โ ๊ฒฐ๊ณผ๋ฌผ์ ๋ด๋๊ธฐ๋ ํ๋ค๊ณ ๋ง๋ถ์์ต๋๋ค.
๋๋น: โ๋ฐ์ด๋ธ ์ฝ๋ฉโ์ ์๋ก์ด ์งํ์ ์ฐ โ์ฌ๋ผ์ด๋โ
Monologue์ GM์ธ ๋๋น(Naveen)์ GPT-5.5์ ๋ํ ๊ฐ์ฅ ์ด์ ์ ์ธ ์ง์ง์ ์ค ํ ๋ช ์ ๋๋ค. ๊ทธ๋ ๊ณผ๊ฑฐ์๋ ์น ์ฑ ์ฝ๋๋ ๊ธ์ฐ๊ธฐ ์์ ์ Claude/Opus๋ฅผ ์ฌ์ฉํ์ง๋ง, GPT-5.5 ์ถ์ ์ดํ์๋ ๋ค๋ฅธ ๋ชจ๋ธ์ ์ฐพ์ ํ์๊ฐ ์๋ค๊ณ ๋งํ์ต๋๋ค. Python, Swift, macOS ๋ค์ดํฐ๋ธ ์ฑ ๊ฐ๋ฐ, ์ฌ์ง์ด ๊ณ ๊ฐ ์ง์ ๋ต๋ณ ์์ฑ์ ์ด๋ฅด๊ธฐ๊น์ง ๋ชจ๋ ๋ฉด์์ โ๋งค์ฐ ์ ํ๋ จ๋(well-rounded)โ ๋ชจ๋ธ์ด๋ผ๊ณ ๊ทน์ฐฌํ์ต๋๋ค.
ํนํ ๋๋น์ GPT-5.5๊ฐ โ๋ฐ์ด๋ธ ์ฝ๋ฉ(Vibe Coding)โ์ ์ง์ ํ ๊ฐ์๋ผ๊ณ ๊ฐ์กฐํ์ต๋๋ค. ๋ฐ์ด๋ธ ์ฝ๋ฉ์ ์์ธํ ๊ณํ ์์ด ์์ด๋์ด๋ฅผ ๋ฐ๋ก ์ฝ๋๋ก ๊ตฌํํ๋ ๋ฐฉ์์ ์๋ฏธํฉ๋๋ค. ๊ทธ๋ ์ง๋ ๋ช ์ฃผ๊ฐ GPT-5.5๋ฅผ ์ด์ฉํด โDaylineโ์ด๋ผ๋ Raycast ๋์ฒด ์ฑ์ ํฌํจํ์ฌ ์ธ ๊ฐ์ง ์ฑ์ ์ฑ๊ณต์ ์ผ๋ก ๋ฐ์ด๋ธ ์ฝ๋ฉํ๋ค๊ณ ๋ฐํ์ต๋๋ค. ๋๋๊ฒ๋ Dayline ์ฑ์ ๋จ์ผ ์ค๋ ๋์์ 2์ต ํ ํฐ(200 million tokens)์ด๋ผ๋ ์์ฒญ๋ ์์ ์ปจํ ์คํธ๋ฅผ ์ ์งํ๋ฉฐ ๊ฐ๋ฐ๋์์ผ๋ฉฐ, ๊ทธ๋ ๋จ ํ ์ค์ ์ฝ๋๋ ์ง์ ์์ฑํ์ง ์๊ณ ์คํฌ๋ฆฐ์ท๊ณผ ์์ด๋์ด๋ง์ผ๋ก ์ฑ์ ์์ฑํ ์ ์์์ต๋๋ค.
๋๋น์ ๊ฒฝํ์ GPT-5.5๊ฐ ์ฌ๋ฌ ์ฝ๋๋ฒ ์ด์ค์ ์ปจํ ์คํธ๋ฅผ ๋์์ ๊ด๋ฆฌํ๋ ๋ฐ ํ์ํ๋ฉฐ, โ์์ถ(compaction)โ ๊ธฐ์ ์ ํตํด ๊ธด ๋ํ ์ค๋ ๋์์๋ ์ปจํ ์คํธ๋ฅผ ์์ง ์๋ ๋ฅ๋ ฅ์ด ๋ฐ์ด๋๋ค๋ ๋์ ๋ถ์๊ณผ ์ผ์นํฉ๋๋ค. ์ด๋ Opus๊ฐ ๋๊ท๋ชจ ํ๋ก์ ํธ์์ โ๊ฒ๋จน๋(intimidated)โ ๊ฒ๊ณผ ๋์กฐ์ ์ผ๋ก, GPT-5.5๊ฐ ์ฃผ์ด์ง ์ธ๋ถ ์ ๋ณด๋ฅผ ๋จธ๋ฆฟ์์ ๋ด๊ณ ์ฒ์๋ถํฐ ๋๊น์ง ์์ ํ ๊ฒฐ๊ณผ๋ฌผ์ ๋ง๋ค์ด๋ด๋ ๋ฅ๋ ฅ์ ๋ณด์ฌ์ค๋๋ค.
์ค์คํด: ์ง์ ์ ๋ฌด์ ํ์ , ์๋๊ฐ ๊ณง ์ง๋ฅ์ด๋ค
Every์ ์ฑ์ฅ ์ฑ ์์(Head of Growth)์ธ ์ค์คํด(Austin)์ GPT-5.5๊ฐ ๊ทธ์ ์ง์ ์ ๋ฌด ๋ฐฉ์์ ์์ ํ ๋ฐ๊พธ์ด ๋์๋ค๊ณ ๋งํ์ต๋๋ค. ๊ทธ๋ ๊ธฐ์ ์ ๋ฐฐ๊ฒฝ์ด ์์์๋ ๋ถ๊ตฌํ๊ณ , Codex ์ฑ์ ํตํด GPT-5.5๋ฅผ ์ฌ์ฉํ์ฌ ๋์๋ณด๋ ๊ตฌ์ถ, ๋๋ฉ ํ์ด์ง ๊ฐ๋ฐ, ๋ฐ์ดํฐ ๋ถ์, ์บ ํ์ธ ๊ณํ ์๋ฆฝ ๋ฑ ๋ค์ํ ์์ง๋์ด๋ง ์์ ์ ์ํํ ์ ์๊ฒ ๋์์ต๋๋ค.
๊ณผ๊ฑฐ OpenAI ๋ชจ๋ธ๋ค์ ๊ทธ์๊ฒ โ๋ฉ์ฒญํ๋ค๋ ๋๋(makes me feel dumb)โ์ ์ฃผ์์ง๋ง, GPT-5.5๋ ๋ ์ด์ ๊ทธ๋ ์ง ์๋ค๊ณ ํฉ๋๋ค. ๋ชจ๋ธ์ โ์น๊ทผํจ(personable)โ์ด ํฌ๊ฒ ๊ฐ์ ๋์๊ณ , ์ง๋ฌธ์ ๋ช ํํ๊ฒ ์ ์ํ๋ฉฐ, ๊ทธ๊ฐ ์ ์ํ ๋งฅ๋ฝ์ ๋ฐํ์ผ๋ก ์๋ฏธ ์๋ ๋ต๋ณ์ ๋ด๋๋๋ค๊ณ ์ค๋ช ํ์ต๋๋ค.
๊ฐ์ฅ ์ธ์์ ์ธ ์ฌ๋ก๋ก, ๊ทธ๋ ๊ณง ์ถ์๋ โํ๋ฌ์ค ์(plus one)โ ์ ํ์ ์บ ํ์ธ ๊ณํ์ GPT-5.5์ ๋งก๊ฒผ์ต๋๋ค. Notion, Slack, ๊ธฐ์กด ์บ ํ์ธ ๊ณํ ๋ฑ ๋ชจ๋ ๊ด๋ จ ์๋ฃ๋ฅผ ๋ชจ๋ธ์ ์ ๊ณตํ๊ณ ๋๋ต์ ์ธ ๋ชฉํ๋ฅผ ์ ์ํ์, GPT-5.5๋ ํ์ ์ทจํฅ๊ณผ ๋ ผ์ ๋ด์ฉ์ ๋ฐ์ํ ์บ ํ์ธ ๊ณํ์ 90% ์์ฑ๋ ์๊ฒ ์ ์ํ์ต๋๋ค.
์ค์คํด์ GPT-5.5์ โ์๋โ๊ฐ ํต์ฌ์ ์ธ ๊ฐ์ ์ด๋ผ๊ณ ๊ฐ์กฐํ์ต๋๋ค. ๊ทธ๋ ํ์ ์ค์ด๋ NBA ํ๋ ์ด์คํ ๊ฒฝ๊ธฐ๋ฅผ ์์ฒญํ๋ ์ค์๋ ๋์ 10%๋ง ์ฌ์ฉํ๋ฉฐ ๋ชจ๋ธ๊ณผ ์ํธ์์ฉํ ์ ์์ ์ ๋๋ก GPT-5.5๊ฐ ๋น ๋ฅด๊ฒ ์์ ์ ์ฒ๋ฆฌํ๋ค๊ณ ๋งํ์ต๋๋ค. ๊ทธ๋ โ์๋๋ ์ด๋ค ๋ฉด์์๋ ์ง๋ฅ์ ํ ์ข ๋ฅ(speed is in some ways a type of intelligence)โ๋ผ๊ณ ๋ง๋ถ์ด๋ฉฐ, GPT-5.5์ ๋น ๋ฅธ ์๋๊ฐ ์ฌ์ฉ์์๊ฒ ์์ฒญ๋ ๊ถํ์ ๋ถ์ฌํ๋ค๊ณ ์ญ์คํ์ต๋๋ค.
GPT-5.5์ ํน์ง ๋ฐ ์์ฌ์
GPT-5.5๋ ๋จ์ํ ์ฑ๋ฅ์ด ํฅ์๋ ๋ชจ๋ธ์ ๋์ด, AI์์ ์ํธ์์ฉ ๋ฐฉ์๊ณผ ์์ ํ๋ฆ์ ๊ทผ๋ณธ์ ์ธ ๋ณํ๋ฅผ ๊ฐ์ ธ์ค๊ณ ์์ต๋๋ค.
- ์๋์ ์ง๋ฅ์ ๊ฒฐํฉ: GPT-5.5์ ์๋์ ์ธ ์ฒ๋ฆฌ ์๋๋ ์ฌ์ฉ์๊ฐ ๋ ๋ง์ ์์ ์ ๋์์ ์ํํ๊ณ , ์์ด๋์ด๋ฅผ ๋น ๋ฅด๊ฒ ์คํํ๋ฉฐ, ์ง์ ์ ๋ฌด์ ํจ์จ์ฑ์ ๊ทน๋ํํ ์ ์๋๋ก ๋์ต๋๋ค.
- ์ปจํ ์คํธ ๊ด๋ฆฌ ๋ฅ๋ ฅ: ์์ต ํ ํฐ์ ๋ฌํ๋ ๊ธด ๋ํ ์ค๋ ๋์์๋ ์ปจํ ์คํธ๋ฅผ ์์ง ์๊ณ ๋ณต์กํ ํ๋ก์ ํธ๋ฅผ ์ฒ์๋ถํฐ ๋๊น์ง ์คํํ๋ ๋ฅ๋ ฅ์ ๋๊ท๋ชจ ๊ฐ๋ฐ ํ๋ก์ ํธ๋ ์ฅ๊ธฐ์ ์ธ ๊ธฐํ์ ํ๋ช ์ ์ธ ๋ณํ๋ฅผ ๊ฐ์ ธ์ฌ ์ ์์ต๋๋ค.
- ๊ณํ๊ณผ ์คํ์ ๋ถ๋ฆฌ: Opus 4.7์ ๋ฐ์ด๋ ๊ณํ ์๋ฆฝ ๋ฅ๋ ฅ๊ณผ GPT-5.5์ ๊ฐ๋ ฅํ ์คํ๋ ฅ์ด ๊ฒฐํฉ๋ ๋ ์ต๊ณ ์ ์๋์ง๋ฅผ ๋ฐํํ๋ค๋ ์ ์, ํฅํ AI ๋ชจ๋ธ ๊ฐ์ ํ์ ์ํ๊ณ๊ฐ ๋์ฑ ์ค์ํด์ง ๊ฒ์์ ์์ฌํฉ๋๋ค.
- ๋ค์ํ ์ฌ์ฉ์ ํ๋กํ์ผ์ ๋ํ ์ ํฉ์ฑ: GPT-5.5๋ โ์์ง๋์ด ์์ง๋์ดโ์ ๊ฐ์ ํน์ ๊ธฐ์ ์์ญ์ ๊น์ด ๊ด์ฌํ๋ ์ฌ์ฉ์์๊ฒ๋ ์ต๊ณ ์ ์ ํ์ด ๋ ์ ์์ง๋ง, โ์ ํ ์์ง๋์ด ์ ๋๋ด๋ฆฌ์คํธโ์ ๊ฐ์ด ๊ด๋ฒ์ํ ์ ๋ฌด๋ฅผ ์ํํ๋ ์ฌ์ฉ์์๊ฒ๋ ์ฌ์ ํ Claude์ ๊ฐ์ ์ ๋๋ด๋ฆฌ์คํธ ๋ชจ๋ธ์ด ๋ ์ ํฉํ ์ ์์ต๋๋ค.
- ์๋ก์ด ์ด๋ฏธ์ง ์์ฑ ๋๊ตฌ์ ๋ฑ์ฅ: GPT-5.5์ ํจ๊ป ๊ณต๊ฐ๋ OpenAI์ ์๋ก์ด ์ด๋ฏธ์ง ์์ฑ ๋๊ตฌ๋ ์ ํ๋ธ ์ธ๋ค์ผ๊ณผ ๊ฐ์ ๊ณ ํ์ง์ ์๊ฐ์ ์ฝํ ์ธ ๋ฅผ ๋น ๋ฅด๊ณ ์ฝ๊ฒ ์์ฑํ ์ ์๊ฒ ํ๋ฉฐ, ๋์์ธ ์์ ์์ญ์์๋ AI์ ์ํฅ๋ ฅ์ด ํ๋๋ ๊ฒ์์ ์๊ณ ํฉ๋๋ค.
๊ฒฐ๋ก : AI ํ์ ์ ์๋ก์ด ์๋
GPT-5.5๋ ์์ง๋์ด๋ง ์คํ ๋ฅ๋ ฅ๊ณผ ์ง์ ์ ๋ฌด ํจ์จ์ฑ ๋ฉด์์ AI ๋ชจ๋ธ์ ์๋ก์ด ๊ธฐ์ค์ ์ ์ํ์ต๋๋ค. ํนํ, ๋๊ท๋ชจ ์ฝ๋ ์ฌ์์ฑ ํ๋ก์ ํธ๋ฅผ ์ถ์งํ๋ โ์ฉ๊ธฐโ์ ์์ต ํ ํฐ์ ๋ฌํ๋ ์ปจํ ์คํธ๋ฅผ ์ ์งํ๋ฉฐ โ๋ฐ์ด๋ธ ์ฝ๋ฉโ์ ๊ฐ๋ฅํ๊ฒ ํ๋ ๋ฅ๋ ฅ์ ๊ฐ๋ฐ์๋ค์ ์์ ๋ฐฉ์์ ๊ทผ๋ณธ์ ์ผ๋ก ๋ณํ์ํฌ ์ ์ฌ๋ ฅ์ ๊ฐ์ง๊ณ ์์ต๋๋ค.
๋ฌผ๋ก , Opus 4.7์ ๋ฐ์ด๋ ๊ณํ ๋ฅ๋ ฅ์ด๋ Claude์ ์ ๋๋ด๋ฆฌ์คํธ์ ๊ฐ์ , ๊ทธ๋ฆฌ๊ณ ์ฌ์ ํ ๋์์ธ์ด๋ ํน์ ์ธ์ด(์: Ruby) ์ง์์์ ๋ฏธ๋ฌํ ์ฐจ์ด๊ฐ ์กด์ฌํฉ๋๋ค. ํ์ง๋ง ์ด์ ์ฐ๋ฆฌ๋ ๊ฐ์์ ํ์์ ๋ฐ๋ผ ์ต์ ์ AI ๋ชจ๋ธ์ ์ ํํ๊ณ , ์ฌ์ง์ด ์ฌ๋ฌ ๋ชจ๋ธ์ ์กฐํฉํ์ฌ ํ์ฉํ ์ ์๋ โํธํ๋ก์ด(luxury)โ ์๋์ ์ด๊ณ ์์ต๋๋ค.
Every ํ์ โ๋ฐ์ด๋ธ ์ฒดํฌโ๋ AI ๋ชจ๋ธ์ ๋ฐ์ ์ด ๋จ์ํ ๊ธฐ์ ์ ์ฑ๋ฅ ํฅ์์ ๋์ด, ์ธ๊ฐ-AI ํ์ ์ ๋ณธ์ง๊ณผ ์ญํ ๋ถ๋ด์ ๋ํ ๊น์ ์ง๋ฌธ์ ๋์ง๊ณ ์์์ ๋ณด์ฌ์ค๋๋ค. GPT-5.5๋ ์ด๋ฌํ ํ์ ์ ๋ค์ ๋จ๊ณ๋ฅผ ์ํ ๊ฐ๋ ฅํ ๋๊ตฌ์ด๋ฉฐ, ์์ผ๋ก ์ธ๊ฐ๊ณผ AI๊ฐ ์ด๋ป๊ฒ ์ํธ์์ฉํ๋ฉฐ ์๋ก์ด ๊ฐ์น๋ฅผ ์ฐฝ์ถํด ๋๊ฐ์ง ๊ธฐ๋ํ๊ฒ ๋ง๋ญ๋๋ค.