Skip to content
Latest

What is moving in AI, read closely

We read the papers, the release notes and the research programmes so you do not have to. Every post is dated, sourced and checked line by line. Start at the top of the reading path if you are new; jump to the frontier if you are not.

The highlighted you should know points in each post are the parts most used in industry right now. If you remember nothing else, remember those.
model historytimelinefrontier models 14 min

From GPT-1 to today: what actually changed in eight years of models

A dated walk through the model releases that changed how these systems are built, priced and deployed, from June 2018 to September 2026.

Eight years separate GPT-1's 117 million parameters and 512-token context from GPT-6 Astra's 1.05 million-token context. In between, four things changed the shape of the field: pre-training at scale, instruction tuning with human feedback, reinforcement learning for chain-of-thought reasoning, and a standard agent stack. Move the PaLM entry (5 April 2022) after the InstructGPT entry (4 March 2022) so the timeline actually runs in date order; leave this sentence as written.

you should know

Three recipe changes carried the field, not one: generative pre-training (GPT-1, June 2018), instruction tuning with human feedback (InstructGPT, March 2022), and reinforcement learning for chain-of-thought reasoning (o1, September 2024).

Sep 11, 2026Read
architecturetransformersinference 12 min

The architecture story: from the transformer to reasoning and agents

Nine ideas, grouped by what each one changed for people building products on top of these models.

Modern language models are the result of a sequence of separable ideas: the transformer, scaling laws, post-training, sparse experts, long context, inference efficiency, reinforcement learning for reasoning, multimodality and tool use, and finally protocols. This post walks the sequence in order, with dates and numbers from the cited sources, and states the practical consequence of each step for anyone shipping a product.

you should know

Parameter count no longer predicts serving cost. Sparse mixture-of-experts models activate a fraction of their weights per token: DeepSeek-V3 is 671B total but 37B active, and Mixtral 8x7B was 47B total with 13B active.

Sep 11, 2026Read
evalsretrievalfine-tuning 11 min

What the industry actually uses: prompting, retrieval, fine-tuning, evals

The techniques teams reach for, in the order they reach for them, with the survey numbers behind each one.

Enterprise generative-AI spend hit roughly $37 billion in 2025, but the technique list barely moved. Prompt design is still the dominant customisation method, retrieval is second, 57% of teams fine-tune nothing at all, and only 52.4% run offline evals. This post walks the stack in the order teams actually adopt it, with the numbers and dates for each step.

you should know

Prompt design is still the most widely used LLM customisation technique in enterprises, with retrieval second and fine-tuning niche (Menlo Ventures, December 2025).

Sep 11, 2026Read
evalsbenchmarksllm-as-judge 14 min

Evals: the discipline that separates demos from products

How teams that ship LLM systems actually measure them: golden sets, graders, judge bias, saturating benchmarks and release gates.

Evaluation has become a named engineering function, with published practice from both Anthropic and OpenAI. This post covers how to build a golden set, which graders to use, how LLM-as-judge fails, why agentic benchmarks are saturating, and what a release gate should look like. It ends with the practical question of who owns your eval data when a vendor shuts a product down.

you should know

Start with 20-50 tasks drawn from real failures - bug trackers, support queues, spot checks - and prove each is solvable by writing a reference solution first (Anthropic, 9 January 2026).

Sep 11, 2026Read
interpretabilityauditingAI safety 11 min

Mechanistic interpretability: reading what a model is doing, and why businesses will care

A plain-language account of features, sparse autoencoders, circuits, steering and chain-of-thought monitoring, and what each one can and cannot do for an audit.

A language model is a pile of matrices, and when it approves, drafts or refuses something, nothing in the system tells you why. Mechanistic interpretability is the effort to reverse-engineer those matrices into readable algorithms, and over four years it has produced real results: sparse autoencoders that name concepts, attribution graphs that trace a prompt to an output, steering vectors that adjust behaviour below the prompt layer, and chain-of-thought monitors that catch reward hacking. This post walks the landmark results in order and says plainly what each one buys you when you have to audit or debug a deployed system.

you should know

Individual neurons mostly do not mean one thing: Anthropic's Toy Models of Superposition (21 September 2022) showed networks pack more concepts than they have dimensions, which is why single-neuron explanations fail.

Sep 11, 2026Read
alignmentAI safetyagents 13 min

Alignment research you should know: OpenAI, Anthropic, MATS and Berkeley

A survey of the main alignment research programs and the landmark papers behind them, and what each one implies for a company putting agents into production.

Alignment research now has a settled institutional shape: a talent pipeline at MATS, an academic anchor at Berkeley's CHAI, independent labs like Redwood, Apollo and METR, a state funder in the UK AI Security Institute, and safety teams inside OpenAI and Anthropic. This post walks through the landmark papers those groups produced between December 2022 and July 2026, and says plainly what each one changes for a company deploying agents.

you should know

Alignment properties are global, not task-local: fine-tuning GPT-4o narrowly on insecure code produced broad misalignment on unrelated prompts (Betley et al., 24 February 2025).

Sep 11, 2026Read
agentsarchitectureprotocols 10 min

Agent architectures that real teams ship

The patterns, frameworks and protocols that are actually running in production, with the costs, the failure modes and the numbers behind them.

Most production agent work still runs on a small vocabulary set out in December 2024: five workflow patterns and one definition of an agent. Since then the field has added hard cost numbers, a published taxonomy of 14 multi-agent failure modes, two donated protocols and a 2026 shift from prompting to harness engineering. This post collects what is documented, with dates and figures, and what it implies for teams choosing an architecture.

you should know

Anthropic's "Building effective agents" (19 December 2024) splits systems into workflows, where LLMs are orchestrated through predefined code paths, and agents, where the model directs its own process; it names five workflow patterns still used as the default vocabulary.

Sep 11, 2026Read
DatabricksReplitagent platforms 11 min

How Databricks and Replit build AI in 2026, and how we plug into them

What each platform actually ships today, how customers use it, and where Tachyon connects the two.

Databricks and Replit are the two platforms that come up most often when clients ask where their AI work should live. This post sets out what each one shipped through September 2026, with dates, how customers actually use them, and how Tachyon integrates with each depending on whether the constraint is governed data or delivery speed. It ends with the guardrails we add on top of both.

you should know

Databricks renamed its AI products into the core platform: the platform page is now the 'Databricks Data + AI Platform' with the tagline 'Accelerate the agentic enterprise', Mosaic AI Vector Search is now Databricks AI Search, and 2024-2025 documentation you find online will use stale names.

Sep 11, 2026Read
frontier modelsrelease digestagents 12 min

The frontier this month: what shipped and what it means

A dated record of the frontier model and agent releases between June and 11 September 2026, with prices, links and the practical consequence of each.

Between early June and 11 September 2026 every major lab except Mistral shipped a frontier model, and several shipped more than once. This digest lists each release with its date, price, official link and the concrete change it forces on people building systems. Where sources disagree on a number, both are printed; where a claim could not be verified, it is marked as such.

you should know

GPT-6 Astra shipped to approved users on 3 September 2026 and generally on 4 September at $10/$50 per million tokens with a 1.05M context, but any request over 272,000 input tokens bills at 2x input and 1.5x output for the entire request.

Sep 11, 2026Read

Let's build intelligent systems that drive growth

Tachyon is the engineering partner for teams that need AI in production, not in a deck. Start with a free 60-minute discovery call.

Book a discovery callSend us a brief