<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>NearSync blog: NearSync news</title>
    <link>https://nearsync.ai/blog/news/</link>
    <atom:link href="https://nearsync.ai/blog/news/rss.xml" rel="self" type="application/rss+xml" />
    <description>Company news, launches and NearSync for Startups.</description>
    <language>en</language>
    <lastBuildDate>Mon, 16 Nov 2026 05:00:53 GMT</lastBuildDate>
    <item>
      <title>Product Update, GPT-6 Astra and Fable 5.1 now available</title>
      <link>https://nearsync.ai/blog/product-update-gpt-6-astra-and-fable-5-1-now-available/</link>
      <guid isPermaLink="true">https://nearsync.ai/blog/product-update-gpt-6-astra-and-fable-5-1-now-available/</guid>
      <pubDate>Mon, 07 Sep 2026 16:43:17 GMT</pubDate>
      <dc:creator>Shafaf Bhat</dc:creator>
      <category>NearSync news</category>
      <description>Three new AI models are live across NearSync. What each one is for, what it costs in credits, and why your default did not change.</description>
      <content:encoded><![CDATA[<p>The September model catalogue is live. Three new models are selectable across
NearSync, and every one of them now tells you what it costs in credits before you
choose it.</p>
<p>This post explains what is available, how to choose, and what each one draws from
your allowance.</p>
<h2>Which AI models are new</h2>
<p><strong>Gemini 3.8 Flash</strong> is now the recommended Google model. A one million token
context window, and the lightest model in the catalogue on credits.</p>
<p><strong>Claude Fable 5.1</strong> is a frontier model with a one million token context window.</p>
<p><strong>GPT-6 Astra</strong> is also frontier, with the same one million token window. Access is
limited by OpenAI rather than by us, so it will not be available to every account.</p>
<p>The catalogue now sorts every model into a tier. Recommended is what we suggest
for everyday work. Premium is a step up in reasoning. Frontier is the most capable
thing each vendor sells. Previous generations stay selectable and are marked
legacy.</p>
<h2>How credits work</h2>
<p>A credit is roughly one question to the assistant. That is measured rather than
estimated, from real usage across more than a thousand requests.</p>
<p>Every seat gets its own allowance each month. Standard is 2,000 credits, Pro is
4,000, and you can top up at any time. Allowances are per seat rather than a
shared pot, so one person running a heavy week does not empty the team's
allowance.</p>
<p>Recorded meetings do not draw on credits. They are bundled by the hour at the
organisation level, and so is email.</p>
<h2>What the new models cost in credits</h2>
<p>A more capable model does more work for each answer, so it draws more credits.
NearSync states that as a multiple of the model we recommend from the same vendor,
because a multiple is a number you can act on and a token price is not.</p>
<p>Against their vendor's recommended model, per answer:</p>
<table>
<thead>
<tr>
<th>Model</th>
<th>Tier</th>
<th>Credits per answer</th>
</tr>
</thead>
<tbody><tr>
<td>Gemini 3.8 Flash</td>
<td>recommended</td>
<td>the Google baseline</td>
</tr>
<tr>
<td>Claude Sonnet 5</td>
<td>recommended</td>
<td>the Anthropic baseline</td>
</tr>
<tr>
<td>GPT-5.6 Terra</td>
<td>recommended</td>
<td>the OpenAI baseline</td>
</tr>
<tr>
<td>Claude Opus 5</td>
<td>premium</td>
<td>about 2.5x Sonnet 5</td>
</tr>
<tr>
<td>Claude Fable 5.1</td>
<td>frontier</td>
<td>about 5x Sonnet 5</td>
</tr>
<tr>
<td>GPT-6 Astra</td>
<td>frontier</td>
<td>about 4.2x GPT-5.6 Terra</td>
</tr>
</tbody></table>
<p>In practice, on a Standard seat with 2,000 credits, that is roughly two thousand
questions a month on a recommended model, or around four hundred on a frontier
one. Same allowance, different answer to how far it goes.</p>
<h2>How to choose an AI model</h2>
<p>For most of what a business asks software to do, the recommended model is the
right answer. Summarising a thread, drafting a reply, pulling a figure out of a
document, classifying an email: these are not hard problems for a current model,
and a frontier model does not answer them better. It just answers them for five
times the credits.</p>
<p>Frontier models earn their place on long reasoning and on code. A complex
multi-step analysis across a large document set is where the difference is real
and worth paying for.</p>
<h2>Why the context window matters</h2>
<p>All three new models carry a one million token context window. Several older ones
in the catalogue are 128,000 or 200,000.</p>
<p>The context window is how much material a model can hold in view at once. For a
short question it is irrelevant. It becomes the deciding factor when the work is
large: reading a full contract alongside the thread that negotiated it, or
reasoning across a year of invoices rather than a sample.</p>
<p>It is a separate thing from capability, and worth knowing, because a recommended
model with a large window often beats a more capable model with a small one on
exactly the tasks people assume need the expensive option.</p>
<h2>Why we show what each model costs</h2>
<p>Every premium and frontier model carries a line stating its cost per answer
against the recommended model from the same vendor. Option labels say it too:
premium models are marked high usage, frontier models very high usage.</p>
<p>That appears wherever a model is chosen, underneath the active model in the
Control Centre, and again where a department's model is set. If you select the
most capable option we offer, you are told what it draws before you commit to it,
and told again while it is running.</p>
<p>We know this makes our own top-tier options look expensive. A picker that lists
eighteen models sorted by how new they are will move more frontier usage than one
that writes the multiple beside them. The alternative is somebody meeting that
number for the first time when their allowance runs out mid-month.</p>
<h2>Your default AI model did not change</h2>
<p>Adding three models changed no organisation's default.</p>
<p>The easier version of a catalogue update moves everyone to the newest model, on
the reasoning that newer is better and most people never open the setting. It
would have improved some answers. It would also have burned through allowances
nobody agreed to spend and changed the behaviour of every automation running
against the previous model, mid-month, without notice.</p>
<p>Previous generations stay exactly where they are, still selectable, marked legacy
so you can see they are a generation behind. Moving is a choice you make.</p>
<h2>Why some models are limited access</h2>
<p>Tier tells you what a model draws. It does not tell you whether you can have it.</p>
<p>Most models are generally available. Gemini 3.1 Pro is a preview, which means the
vendor can change or withdraw it. GPT-6 Astra is limited, which means OpenAI
decides who gets access, and a key without that access is refused.</p>
<p>We list limited models rather than hiding them until everyone can use one, and the
picker says so plainly.</p>
<h2>Where to change your AI model</h2>
<p>Model selection lives in Intelligence HQ. Your organisation default sits in the
Control Centre, and routing lets you put a specific department on a specific
model. Both read the same catalogue, so a model added centrally appears
everywhere at once. Changing the model is an admin action.</p>
<p>Routing per department is worth using. Finance and Support ask different things of
a model, and there is no reason both should sit on the same one. A department
doing document-heavy reasoning can run on a premium model while everything else
stays on the recommended one, so the higher rate applies only where it earns it.</p>
<p>If you are not sure which tier your work needs, start on the recommended model. It
is the lightest on credits and it handles more than people expect.</p>
]]></content:encoded>
    </item>
    <item>
      <title>We removed the colour amber from our software. Here is why.</title>
      <link>https://nearsync.ai/blog/we-removed-the-colour-amber-from-our-software/</link>
      <guid isPermaLink="true">https://nearsync.ai/blog/we-removed-the-colour-amber-from-our-software/</guid>
      <pubDate>Thu, 13 Aug 2026 07:00:00 GMT</pubDate>
      <dc:creator>Shafaf Bhat</dc:creator>
      <category>NearSync news</category>
      <description>Six decisions from building NearSync, each written with what it cost us. Removing the colour amber from every screen was only the first.</description>
      <content:encoded><![CDATA[<p>Software has a colour it uses to make you anxious, and it is amber. The badge that says fourteen things need you. The banner that says storage is nearly full. The countdown that turns yellow as a renewal date gets close. It works, which is exactly why everyone uses it.</p>
<p>There is no amber in NearSync. Not toned down. Removed.</p>
<p>On its own that is a small decision. But it was the first of a series, and the pattern only shows when they sit side by side. So here are six decisions we took while building NearSync: what we decided, what it looks like in the product, and what it cost us. The costs are the honest part. Anyone can publish a list of values. The bill is what proves you paid for them.</p>
<h2>1. No amber, anywhere</h2>
<p>In NearSync, something genuinely broken gets a small rose dot. Everything else stays quiet. We allow ourselves exactly one flash of amber in the entire product: a warning toast that appears, says its piece, and dismisses itself. Nothing amber is permitted to sit on your screen and hum at you.</p>
<p>We did not start here. Early dashboards had traffic lights, the way almost every dashboard does. Taking amber out meant redrawing every dashboard, every chart palette and every severity scale around a simpler question: is this fine, or does it need a person. The middle state, the amber state, had to become words instead of a colour, because words can be read calmly and a yellow triangle cannot.</p>
<p><strong>What it cost.</strong> Amber is the software industry's pressure lever. It hurries renewals, nudges upgrades, and inflates the feeling that a tool is indispensable. We gave that lever up, and we paid twice: weeks of design rework at the time, and an ongoing price ever since, because we cannot paint a deadline yellow to make you act on it. We think the trade is right. If someone spends eight hours a day inside a screen, that screen should not be allowed to raise their pulse for its own benefit.</p>
<h2>2. Leave actually means leave</h2>
<p>When leave is approved in NearSync, the person's calls stop ringing and their notifications arrive silently until they are back. Nobody toggles anything and nobody has to remember. The system knows they are away, because the leave request lives in the same system as the phone that would have rung.</p>
<p><strong>What it cost.</strong> Every communications product loves the sentence "reach anyone, any time", and we cannot say it. Buyers have asked whether a manager can override the quiet for something urgent, and the answer is no. Not every buyer has liked that answer. It also cost real engineering: leave had to be threaded through the calling stack and every notification rule, weeks of work that shows up on a feature list as nothing at all, because the feature is an absence.</p>
<h2>3. Breaks are part of the work</h2>
<p>The home screen of NearSync greets you with rings, and one of them is Vitality: did you take your break, did you book focus time, did you speak to a colleague today. It sits beside the rings that track output, at the same size, in the same place, every day.</p>
<p><strong>What it cost.</strong> The home screen is the most expensive space in any product, and every department would like a piece of it. We spent a permanent share of it on a measure that no procurement checklist has ever asked for, and to a certain kind of buyer it reads as soft. We kept it anyway. The alternative is a home screen that says the only interesting thing about you is your output, and we do not believe that.</p>
<h2>4. Nobody is monitored</h2>
<p>There is no screen recording in NearSync, no keystroke counting, no screenshots on a timer. Not switched off by default. Not available. We measure work by what got done: the deals that moved, the invoices that went out, the tickets that closed. The same rule holds inside our own company. We are remote first, nobody watches anybody work, and productivity is the only thing we track.</p>
<p><strong>What it cost.</strong> Employee monitoring is a real market with real budgets, sitting directly beside everything we sell, and buyers do ask for it. Each time, we are declining revenue on purpose. There is an engineering cost too, and it points the wrong way round: a screenshot timer is a week of work, while showing output honestly takes a system where the work actually lives. We built the harder one.</p>
<h2>5. One price, and nothing held back</h2>
<p>Every department in NearSync ships in every plan. Sales, finance, people, projects, support, all of it. The difference between our $25 seat and our $35 seat is AI usage, 2,000 credits against 4,000, where one credit is one question to the AI. That is the whole difference. There is no feature waiting on the higher tier, and when something new ships, it ships to everyone, including the seat paying the minimum.</p>
<p>We decided this after a year of building, at the moment the draft tier sheet was in front of us. Cutting a finished system into rations felt wrong, so we did not do it.</p>
<p><strong>What it cost.</strong> The expansion ladder. Most software companies grow their revenue by gating features and selling the gate: the upgrade email, the greyed-out button, the tooltip that says a better plan exists. That machinery reliably prints money, and we walked away from it. Our revenue grows only when a customer adds a person or asks the AI more questions, which means it grows only when they are getting more out of the product. We are fine with that being the only way up.</p>
<h2>6. We say the limits out loud</h2>
<p>The compliance badges on our own website say "under way". Not a wall of acronyms, the true state, published. The same habit runs through the product: we state what a thing does not do yet before you have to discover it yourself.</p>
<p><strong>What it cost.</strong> Tenders. A procurement filter that wants the badge this year drops us before a human reads a word. Sales conversations sometimes open with what is not finished, which no salesperson would choose. We believe the cost runs the other way over time, because the customer who buys from an exact company stays, but we will not pretend the invoice never arrived.</p>
<h2>What the six add up to</h2>
<p>Read them together and the pattern is hard to miss. Each decision trades a lever that is known to work, urgency colour, constant reachability, monitoring, feature gates, badge walls, for a system we can look at without flinching. Software teaches people what their company thinks of them. It does this every day, in small print, whether anyone intends it or not. We would rather ours said: your attention is yours, your evenings are yours, your break counts, and what you build here is judged by what it does.</p>
<p>The plain version of what we are building is one system for every department of an emerging company, with AI through all of it that asks before it acts. That is the plan. Underneath the plan is the belief this whole post has been circling: the hours someone spends inside a work system are hours of their life, and the system should behave accordingly.</p>
<p>Everyone near, everything in sync.</p>
]]></content:encoded>
    </item>
  </channel>
</rss>
